Citation
Accuracy of image GPS EXIF data from Apple and Samsung mobile devices compared to GPS unit

Material Information

Title:
Accuracy of image GPS EXIF data from Apple and Samsung mobile devices compared to GPS unit
Creator:
Felix, Daniel
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Master's ( Master of science)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Music and Entertainment Industry Studies, CU Denver
Degree Disciplines:
Recording arts
Committee Chair:
Grigoras, Catalin
Committee Members:
Smith, Jeff M.
Lewis, Jason
Whitecotton, Cole

Notes

Abstract:
Mostly every photo captured from a cellular device contains EXIF GPS coordinates associated to the location of where that photo was captured. This thesis will propose tests for accuracy of GPS EXIF data from different Apple, Samsung and Garmin GPS devices. Apple has used the same GPS satellites in their cellular devices within the past 4 generations of models released. Samsung has used the same GPS satellites from the past 3 generations of models released. Both Apple and Samsung use a total of 4 GPS satellites, but differ in one. The proposed tests will determine if this one different satellite may cause separate results to be produced. Within this experiment two types of tests will be administered. The first test will involve a focus of a cellular device capturing photos by having cell service on, then switching to airplane mode. The second test will involve a focus of a cellular device capturing photos by having airplane mode on initially then switching to cell service active. Results of both tests will be analyzed and any anomalies will be addressed. NGS survey markers will be utilized to explore the idea of a more prominent point established when gathering test information. Image GPS data from cellular devices will be compared to a standalone GPS device readings. Experiments will take place in urban and rural environments and results will be analyzed.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Daniel Felix. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
ACCURACY OF IMAGE GPS EXIF DATA FROM APPLE AND SAMSUNG
MOBILE DEVICES COMPARED TO GPS UNIT
by
DANIEL FELIX
B.S., San Diego State University, 2013
A thesis submitted to the Faculty of Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Recording Arts Program
2019


This thesis for the Master of Science degree by Daniel Felix
has been approved for the Recording Arts Program by
Catalin Grigoras, Chair Jeff M. Smith Jason Lewis Cole Whitecotton
Date: December 14, 2019
11


Felix, Daniel (M.S., Recording Arts Program)
Accuracy of Image GPS EXIF Data from Apple and Samsung Mobile Devices Compared to GPS Unit
Thesis directed by Associate Professor Catalin Grigoras
ABSTRACT
Mostly every photo captured from a cellular device contains EXIF GPS coordinates associated to the location of where that photo was captured. This thesis will propose tests for accuracy of GPS EXIF data from different Apple, Samsung and Garmin GPS devices. Apple has used the same GPS satellites in their cellular devices within the past 4 generations of models released. Samsung has used the same GPS satellites from the past 3 generations of models released. Both Apple and Samsung use a total of 4 GPS satellites, but differ in one. The proposed tests will determine if this one different satellite may cause separate results to be produced.
Within this experiment two types of tests will be administered. The first test will involve a focus of a cellular device capturing photos by having cell service on, then switching to airplane mode. The second test will involve a focus of a cellular device capturing photos by having airplane mode on initially then switching to cell service active. Results of both tests will be analyzed and any anomalies will be addressed. NGS survey markers will be utilized to explore the idea of a more prominent point established when gathering test information. Image GPS data from cellular devices will be compared to a standalone GPS device readings. Experiments will take place in urban and rural environments and results will be analyzed.
The form and content of this abstract are approved. I recommend its publication.
Approved: Catalin Grigoras


This thesis is dedicated to my family. For always supporting me in any aspirations I wanted to
pursue and for guiding me throughout life.
IV


ACKNOWLEDGMENTS
A special thank you for my thesis committee, Catalin, Jeff, Jason and Cole for all the work and guidance towards this thesis. Their support throughout this process was extremely appreciative and brought a great skillset to support this thesis.
Catalin and Jeff, thank you for the past two years throughout the Media Forensics graduate program at CU Denver. You both have showed me skills that I will have the rest of my career and opened my eyes to a new passion of media forensics.
Leah and Cole, thanks for the added support throughout the past two years. You both were easy to approach with any questions that may arise and are an invaluable asset to the Media Forensics program.
To the Media Forensics cohort 2017-19, thanks for all the support throughout our program and I am super grateful do this program with such a great group of individuals.
v


TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION........................................................1
Literature Review......................................................2
II. METHODOLOGY.........................................................4
Description of Materials (Devices Used)................................4
Software...............................................................4
Pic2Map..............................................................5
Google Earth.........................................................5
SARTOPO..............................................................5
JPEGSnoop............................................................5
Matlab...............................................................6
GPS Explanation........................................................6
A-GPS................................................................7
GLONASS..............................................................7
GALILEO..............................................................7
QZSS.................................................................7
BDS..................................................................7
EXIF Photo GPS Explanation from JPEGSnoop..............................8
Method Used to Acquire Data............................................9
vi


Test #1 - Locations #I-IX.....................................................10
Test #11 Locations #X - #XI: NGS Locations and Initial Airplane Mode Tests...11
Analysis to Determine Elevation Error.........................................12
Analysis to Determine Distance from Actual Location..........................13
III. TESTING LOCATIONS AND RESULTS FROM TEST #1 & #11...........................14
Test #1 - Rural Location #1....................................................14
Test #1 - Rural Location #11...................................................17
Test #1 - Rural Location #TTT..................................................20
Test #1 - Rural Location #IV...................................................23
Test #1 - Suburb Location #1....................................................25
Test #1 - Suburb Location #11...................................................27
Test #1 - Suburb Location #111..................................................29
Test #1 - Suburb Location #IV...................................................32
Test #1 - Urban Location #1.....................................................34
Test #11 - NGS Location #1 Urban #11............................................36
Test #11 - NGS Location #11 Rural #V............................................38
IV. FINDINGS AND RESULTS EXPLANATION............................................43
Distance Error..................................................................43
Device........................................................................44
Environment...................................................................44
vii


Rural...............................................................45
Urban...............................................................47
Elevation Error.........................................................48
Device................................................................49
Environment...........................................................50
Apple vs Samsung........................................................50
Airplane Mode...........................................................51
Other Notable Findings..................................................52
V. FUTURE RESEARCH......................................................55
VF CONCLUSION.............................................................57
REFERENCES................................................................60
viii


LIST OF TABLES
TABLE
Table 1: Test #1 - Rural Location #1 Coordinates and Elevation...............................15
Table 2: Rural Location #1 - Percent Error Elevations and Average Distance from Actual.......17
Table 3: Rural Location #11 - Coordinates and Elevation......................................18
Table 4: Rural Location #11 - Percent Error Elevations and Average Distance from Actual......20
Table 5: Rural Location #111 - Coordinates and Elevation.....................................21
Table 6: Rural Location #111 - Percent Error Elevation and Average Distance from Actual......22
Table 7: Rural Location #IV - Rural #IV Coordinates and Elevation............................23
Table 8: Rural Location #IV - Percent Error Elevation and Average Distance from Actual ......25
Table 9: Suburb Location #1 - Coordinates and Elevation......................................25
Table 10: Suburb Location #1 - Percent Error and Average Distance from Actual................27
Table 11: Suburb Location #11 - Coordinates and Elevation....................................28
Table 12: Test #VI Suburb #11 - Percent Error Elevation and Average Distance from Actual ....29
Table 13: Suburb Location #111 - Coordinates and Elevation...................................30
Table 14: Suburb Location #111 - Percent Error Elevation and Distance from Actual............31
Table 15: Suburb Location #IV - Coordinates and Elevation....................................32
Table 16: Suburb Location #IV - Percent Error and Average Distance from Actual...............33
Table 17: Urban Location #1 - Coordinates and Elevation......................................34
Table 18: Urban Location #1 - Percent Error Elevation and Average Distance from Actual.......36
Table 19: NGS Location #1 Urban #11 - Coordinates and Elevation..............................37
Table 20: NGS Location #1 Urban #11 - Percent Error and Average Distance from Actual.........38
Table 21: NGS Location #11 Rural #IV - Coordinates and Elevation.............................40
IX


Table 22: NGS Location #11 Urban #V - Percent Error Elevation and Average Distance from
Actual..................................................................................42
x


LIST OF FIGURES
FIGURE
Figure 1: Devices Used............................................................4
Figure 2: EXIF GPS Example: JPEGSnoop.............................................9
Figure 3: EXIF Hex View (HxD).....................................................9
Figure 4: Photo Example from Cell Phone Device...................................11
Figure 5: Percent Error Formula..................................................13
Figure 6: Rural Location #1 - Anomaly TMG_0119_Airplane.jpg’ Location Compared to Actual
Location (SARTOPO Map)...........................................................15
Figure 7: Rural Location #1 - Image GPS Coordinates vs Actual Location (SARTOPO map)... 16 Figure 8: Rural Location #1 - Image GPS Coordinates vs Actual Location (Google Earth map) 16 Figure 9: Rural Location #11 - Image GPS Coordinates vs Actual Location (SARTOPO map).. 18 Figure 10: Rural Location #11 - Anomaly IMG_1950.jpg Distance from Actual Location
(SARTOPO map)....................................................................19
Figure 11: Rural Location #11 Anomaly #2 - iPhone 7 Plus Distance Corrections to Actual
Location (SARTOPO Map)...........................................................19
Figure 12: Rural Location #111 - Image GPS Coordinates vs Actual Location (SARTOPO Map)
................................................................................22
Figure 13: Rural Location #IV - Image GPS Coordinates vs Actual Location (SARTOPO Map)
.................................................................................23
Figure 14: Rural Location #IV - Image GPS Coordinates vs Actual Location (Google Earth Map).............................................................................24
xi


Figure 15: Suburb Location #1 - Anomaly Samsung Note 9 ‘20190916_102914_Airplane.JPG’
Distance from Actual Location (SARTOPO Map)..........................................26
Figure 16: Suburb Location #1 - Image GPS Coordinates vs Actual Location (SARTOPO Map)
....................................................................................26
Figure 17: Suburb Location #1 - Image GPS Coordinates vs Actual Location (Google Earth
Map).................................................................................27
Figure 18: Suburb Location #11 - Image GPS Coordinates vs Actual Location (SARTOPO Map)
....................................................................................28
Figure 19: Suburb Location #11 - Image GPS Coordinates vs Actual Location (Google Earth
Map).................................................................................29
Figure 20: Suburb Location #111 - Image GPS Coordinates vs Actual Location (SARTOPO Map)
.....................................................................................31
Figure 21: Suburb Location #IV - GPS Coordinates vs Actual Location (SARTOPO Map)....33
Figure 22: Urban Location #1 - Image GPS Coordinates vs Actual Location (SARTOPO Map)35 Figure 23: Urban Location #1 - Image GPS Coordinates vs Actual Location (Google Earth Map)
.....................................................................................35
Figure 24: NGS Location #1 Urban #11 - NGS Data Sheet Rural Environment and Survey marker
54 RESET..................................................................37
Figure 25: NGS Location #1 Urban #11 - Image GPS Coordinates from Actual Location
(SARTOPO Map).............................................................38
Figure 26: NGS Location #11 Urban #V - NGS Data Sheet Urban Environment and Survey Marker RD 2197............................................................39
xn


Figure 27: NGS Location #11 Urban #V - Image GPS Coordinates vs Actual Location
(SARTOPO Map)...........................................................................41
Figure 28: NGS Location #11 Urban #V - Image GPS Coordinates vs Actual Location (Google
Earth Map)..............................................................................42
Figure 29: Distance Error from Actual Location..........................................43
Figure 30: Device Average Distance Pertaining to Environment...........................44
Figure 31: Rural Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph.....46
Figure 32: Suburb Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph....47
Figure 33: Urban Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph.....48
Figure 34: Overall Elevation Percent Error..............................................49
Figure 35 : Elevation Average Percent Error Device Ranking.............................49
Figure 36: Apple vs Samsung Distance from Actual.......................................51
xiii


I.
INTRODUCTION
Photo EXIF data has become popular in the past couple of years to the general public. The illusion of a photo being simply a photo is not the case anymore. The data that can be captured by these image files can encompass personal information of a user. Within a photo, GPS metadata can be captured that contains the exact coordinates of where a person was previously located. This brings up questions of how a cellular device is creating an image file with this background information and the accuracy of this data.
Within this paper, the accuracy of EXIF data being captured from Apple and Samsung cellular devices will be compared to GPS standalone devices. Airplane mode is another feature within cellular devices that will be explored. This feature turns off cell service associated to the device. This function will be experimented on to see if GPS data is still captured and outputted to image metadata. Tests will be conducted in three different environments to determine if accuracy is swayed. Two tests will be done using NGS (National Geodetic Survey) marker locations to see if any different outcomes were noticed.
Tests will be analyzed and compared to determine any interesting trends and findings. These trends could be associated to environment, airplane mode, elevation and phone manufacturer associated to the device. Any anomalies will be addressed and analyzed to determine the reason an anomaly occurred.
This thesis will be organized by chapters and will outline the steps of the overall test process. Chapter I presents the introduction and literature review. Chapter II presents the methodology describing software, devices and methods used in acquiring data. Chapter III presents the testing locations and raw data associated to each location. Chapter IV presents the findings and results
1


from each test. Chapter V presents future research that can be addressed. Chapter VI is the conclusion and key takeaways from this paper.
Literature Review
Previous research was conducted before administering tests to determine a specific methodology.
A thesis titled “Visual Geo-Localization and Location-Aware Image Understanding" by Amir Roshan Zamir was reviewed to see how geo-tags might be shown throughout experiments. Zamirs’ thesis is mainly focused on location aware applications applying a geo tag to images. However, Zamir mentions that many of the procedures which use geo-tags as their input require a precise geo-location, particularly in the urban areas. In respects to other software applications, this data is important to be accurate for the software to run properly. This was acknowledged and the use of different tools will be utilized when processing the data to validate one another. In addition, this paper was helpful to see the different ways of displaying and picking test locations to conduct experiments at.
Another reference was titled “Analysis of errors in EXIF metadata on mobile devices” by Ana Lucila Sandoval Orozco and David Manuel Arenas Gonzalez. This piece was used as reference to see how EXIF metadata in photos might look like and what errors may arise from experiments. Orozco and Gonzalez mention that the area of image forensics analysis can be divided into two large branches: picture authentication and source authentication. Moreover, Orozco and Gonalez states if GPS tags are in place in metadata and display values from 0 to 1, this indicates that the data has a high probability of being wrong. This type of occurrence was not seen in all experiments conducted. However, was taken into account because some data did not capture GPS coordinates and did not display any information whatsoever. Orozco and Gonzalez
2


explain forensic techniques for image analysis, describe image metadata and how this is reviewed.
A reference titled “Smartphone GPS accuracy study in an urban environment’ by Krista Merry was another work that contributed to how the administration of photos will be handled to capture data at different points of interest. Merry explains how phone positioning can hinder the results. Merry states, that at the collection of the first point collected, the phones WIFI capability was disabled. Further, after the collection of the first point the WIFI was enabled and two minutes were allowed to pass before the second data point was collected. While administrating tests, it was confirmed that the tester would take this into consideration and to have devices have a certain time allotted for GPS to be acquired. Merry’s main objective of her study was to determine the accuracy of an iPhone 6s location under GPS only and WIFI only settings.
3


II. METHODOLOGY
Description of Materials (Devices Used)
Devices were used in a series of two tests and eleven locations comparing either an Apple, Samsung or Garmin GPS unit. Figure 1 illustrates the type of device, release date, OS version, network tech, WLAN, Bluetooth and GPS satellites associated to each device. Details from each device is gathered by GSMarena.com, which is a well-known resource for device specifications. It is important to note the type of GPS associated to each device. This can be an indicator that results should vary between each test. For example, Apple devices use a GPS satellite system titled, QZSS, where Samsung utilizes the BDS GPS satellite.
Device Release Date OS Version Network Tech WLAN Bluetooth GPS
Apple iPhone 6s Sept, 2015 12.4.2 GSM/CDMA/HSPA/ EVDO/LTE WiFi 802.11 a/b/g/n/ac, dual band, hotspot 4.2, A2DP, LE A-GPS, GLONASS, GALILEO, QZSS
Apple iPhone 7 Sept, 2016 12.4.2 GSM/CDMA/HSPA/ EVDO/LTE WiFi 802.11 a/b/g/n/ac, dual band, hotspot 4.2, A2DP, LE A-GPS, GLONASS, GALILEO, QZSS
Apple iPhone 7 Plus Sept, 2016 12.4.2 GSM/CDMA/HSPA/ EVDO/LTE WiFi 802.11 a/b/g/n/ac, dual band, hotspot 4.2, A2DP, LE A-GPS, GLONASS, GALILEO, QZSS
Apple iPhone 8 Sept, 2017 12.4.2 GSM/HSPA/LTE WiFi 802.11 a/b/g/n/ac, dual band, hotspot 5.0, A2DP, LE A-GPS, GLONASS, GALILEO, QZSS
Samsung Galaxy S6 March, 2015 7 GSM/HSPA/LTE WiFi 802.11 a/b/g/n/ac, dual band, Wi-FI Direct, hotspot 4.1, A2DP, LE, aptX A-GPS, GLONASS, BDS
Samsung Note 9 Aug, 2018 10 GSM/CDMA/ HSPA / EVDO / LTE WiFi 802.11 a/b/g/n/ac, dual band, Wi-FI Direct, hotspot 5.0, A2DP, LE, aptX A-GPS, GLONASS, BDS, GALILEO
Garmin GPSmap 62s June, 2010 5.3 - - - GPS, WAAS
Garmin eTrex 2 Ox May, 2015 2.00 - - - GPS, GLONASS
Figure 1: Devices Used Software
Multiple types of software were used throughout testing. Many were chosen to validate other tools used. The different type of software and online resources used were Pic2Map, Google Earth, SARTOPO and JPEGSnoop. Below will list each software with a brief description and their function throughout the experiments.
4


Pic2Map
Pic2Map is an online EXIF data viewer with GPS support which allows the user to locate and view your photos on Google maps. Throughout these experiments this tool was utilized for ease of displaying EXIF data within test photos. Also, this tool featured the ability to extract and view GPS coordinates associated to test photos. Pic2Map can be accessed by https://pic2map.co m and from there a user can upload image files for EXIF viewing.
Google Earth
Google Earth is a computer program that renders a 3D representation of Earth based primarily on satellite imagery. This tool was utilized to validate other tools in the testing environment. Google Earth is a great tool for GPS coordinate plotting as it displays each point throughout its updated satellite maps. Google Earth version 7.3 was used.
SARTOPO
SARTOPO is a mapping and trip planning tool for the back country. This tool is widely used in search and rescue operations where it provides simple ease of GPS coordinate plotting. SARTOPO utilizes different type of maps ranging from Google satellite imagery to elevation maps. SARTOPO was a tool mainly used for displaying GPS coordinates and was compared to Google Earth for verification. SARTOPO is an online tool and can be accessed from https://sartopo.com/.
JPEGSnoop
JPEGSnoop is a software that scans the image and offers the user all the detailed information called EXIF data. EXIF data contains information about the camera, edition program, date, color histogram, compression formats and other details associated to the image metadata. JPEGSnoop was used within these experiments as another tool to test data for
5


accuracy and to provide another way for viewing test photos. JPEGSnoop software version 1.7.3 was used.
Matlab
Matlab is a software that combines a desktop environment turned for iterative analysis and design processes with a programming language that expresses matrix and array mathematics directly (https://www.mathworks.com/products/matlab.htmn. Matlab was used through these experiments to calculate the distance between actual known points against coordinates gathered from the test photos. Matlab version 9.7.0.1190202 was used.
GPS Explanation
Understanding a basic idea of how GPS satellites talk to a device is important to grasp a sense of what kind of processes a device may be experiencing in the background. A brief explanation from ‘gps.gov’ explains the type of process most devices on utilize on Earth for acquiring GPS.
GPS is a group of 24 or more satellites flying above the surface of Earth. Each one circles the planet twice a day in one of six orbits to provide continuous, worldwide coverage. GPS satellites broadcast radio signals providing their locations, status, and precise time from on-board atomic clocks. GPS radio signals travel through space at the speed of light, more than 299,792 km/second. GPS devices on Earth receive the radio signals noting their exact time of arrival and use these to calculate its distance from each satellite in view. Once a GPS device knows its distance from at least four satellites, it can use geometry to determine its location on Earth in three dimensions (https://www.gps.gOv/multimedia/poster/y
6


The different satellite systems used throughout the tests are A-GPS, GLONASS, GALILEO, QZSS and BDS. Below is a brief description of each satellite system associated to devices in this experiment that are listed in Figure 1.
A-GPS
A-GPS (Assisted Global Position System) is a procedure GPS chips use to provide accurate positioning. Use of cell service, WIFI and latest GPS system available to provide an location as soon as possible to the device ( https://www.windowscentral.com/gps-vs-agps-quick-tutorial )
GLONASS
GLONASS (Globalnaya Navigazionnaya Sputnikovya Sistema) is a global navigations satellite system owned and operated by the Russian Federations ( https://www.gps.gov/svstem s/gnss/ )
GALILEO
GALILEO is a global navigations satellite system owned and operated by the European Union ( https://www.gps.gov/svstem s/gnss/ )
QZSS
QZSS (Quasi-Zenith Satellite System) is a global navigations satellite system owned by the Government of Japan and operated by the QZS System Service Inc. (QSS). QZSS complements GPS to improve coverage in East Asia and Oceania ( https://www.gps.gov/svstem s/gnss/)
BDS
BDS (BeiDou Navigation Satellite System) is a regional global navigations satellite system owned and operated by the People’s Republic of China. ( https://www.gps.gov/svstem s/gnss/)
7


EXIF Photo GPS Explanation from JPEGSnoop
EXIF photo data is the metadata associated to the photo. Focusing on EXIF GPS data can provide quite a bit of details relating to the image file. Figure 2 is an example of a photo being uploaded to JPEGSnoop and the output referencing GPS EXIF data. JPEGSnoop reads the GPS data associated by its offset in hex relating to the GPS Latitude Ref, GPS Latitude, GPS Longitude Ref, GPS Longitude, GPS Altitude Ref, GPS Altitude, GPS Timestamp, GPS Processing Method and GPS Date Stamp. The alteration of this data cannot be done, as it would change the metadata and ultimately become a new image file. Below is a brief explanation of each type of result JPEGSnoop produces from the GPS EXIF data.
• GPS Latitude Ref gives the direction between ‘North’ and ‘South’ of what coordinates are being captured from the device.
• GPS Latitude displays the latitude of the image file in degree and meters (varies on device and how GPS is being captured)
• GPS Longitude Ref gives the direction between ‘East’ and ‘West’ of what coordinate are being captured from the device.
• GPS Longitude displays the longitude of the image file in degree and meters (varies on device and how GPS is being captured)
• GPS Altitude Ref gives the indication that the elevation is based upon an ‘Above Sea Level’ parameter
• GPS Altitude displays the elevation of the photo being captured usually in meters
• GPS Time Stamp displays the time by hours, minutes and seconds
• GPS Processing Method records the name of the method used for location finding (https://www.exiv2.org/tags-xmp-exifhtmD
8


GPS Date Stamp displays the date of the image file being captured by Year, Month and date

EXIF GPSIFD 0 Absolute 0x0000075C Dir Length = 0x0OOF [GPS Lat i tudeRef [GPSLatitude [GPS Longi tudeRe f [GPSLongitude [GPSAltitudeRef [GPSAltitude [GPSTimeStamp [GPSSpeedRef [GPSSpeed
[GPS ImgDi re cti onRe f [GPS ImgDi re cti on [GPSDestBearingRef [GPSDestBearing [GPSDateStairp
] = ”N”
] = deg 31T 25.970”
] = ”W”
] = deg 40T 49.570" ] = Above Sea Level ] = 13.433 m ] = 1:58:3.00 ] = "km/h"
] = 0.570
] = "True direction"
] = 520198/5165 ] = "True direction"
] = 520198/5165 ] = ”2019:09:20”
Figure 2: EXIF GPS Example: JPEG Snoop
Figure 3 displays the hex view of this information from the given offset seen in information provided by JPEGSnoop. This offset is the start of the EXIF GPSIFD for this image file. Even following the next parameter, GPSLatitudeRef, it can be shown in Figure 3 ACSII view that the latitude reference of ‘N’ is shown. From results within tests administered, EXIF
GPS headers did not display in an image file if that file did not capture any GPS coordinates.
00000730 00 00 00 05 41 70 70 6C 65 00 65 50 63 6F 6E 65 ....Apple.iPhone
00000740 20 33 20 62 61 63 63 20 63 61 6D 65 72 61 20 33 3 hack camera 3
00000750 2E 35 35 6D 6D 20 66 2 F 31 2E 33 00 fK oej 00 01 .55nrai f/1.8.^J. .
00000760 00 02 00 00 00 02 4E 00 00 00 00 02 00 05 00 00 N
00000770 00 03 00 00 08 OA 00 03 00 02 00 00 00 02 57 00 W.
Figure 3: EXIF Hex View (HxD)
Method Used to Acquire Data
In the test phase, two different ways of administering tests were done at eleven different locations. This was due to finding out new information and curiosities that arose from previous tests and locations. Initial Test series #1, focused on the structure order of having cell service active, first photo taken, device switching to airplane mode then second photo taken. This
9


process was working, however was later theorized that the devices could be saving information from previous locations and applying that information to newly created images. In turn of this curiosity, the administration of two more tests were done. Test series #11 was conducted to see if having airplane mode on initially would make a difference in results.
Additionally, actual point locations between Test series #1 was acquired by plotting the point on SARTOPO and making that point the baseline for each test location. Test series #11, introduced the adoption of NGS survey markers, as the actual location parameter to use against the tested image files.
Below are the steps used for the two different types of tests that were administered throughout this experiment.
Test #1 - Locations #I-IX
I. Both cellular devices and GPS units were placed at a stationary position at the location being used as the ‘actual point location’. The tester then waited for the GPS unit to display a GPS coordinate.
II. Then a cellular device was used to take a picture of the GPS unit with cell service on (if cell service was available on the device). This triggers taking a picture of known GPS coordinates with the device creating in the picture the test GPS coordinates.
III. The cellular device was then put in airplane mode under the device settings. Then another photo was taken.
IV. The second GPS standalone device was handled and set in the same location. Steps #II-III were repeated with the cellular device and the second GPS device.
V. With a new cellular device, Steps #II-IV were repeated
VI. Tests concluded once all cellular devices followed steps #II-IV
10


VII. Extraction of images from each cellular device was done by plugging each phone into a computer and extracting images to a USB device. This was to ensure best practices in preserving any metadata associated to the image files.
VIII. Image files were then examined using JPEGsnoop and Pic2map. EXIF data was extracted and organized in Microsoft Excel.
IX. Once all data points were extracted, each point was plotted by latitude and longitude locations from the EXIF data using SARTOPO.
X. Once plotted, maps were extracted from SARTOPO to a .kml file format to view using Google Earth.
Test #11 Locations #X - #XI: NGS Locations and Initial Airplane Mode Tests
I. Cellular device was put into airplane mode and then powered off. Cellular device was then powered back on ensuring airplane mode was still active.
II. GPS unit was placed on top of NGS location marker and tester waited for a GPS coordinate to be displayed on the GPS unit.
11


III. Photo taken from cellular device of GPS unit on top of NGS survey marker. This triggers taking a picture of known GPS coordinates with the device creating in the picture the test GPS coordinates
IV. With a new cellular device, Steps #I-III were administered.
V. Extraction of images from each cellular device was done by plugging each cellular device into a computer and extracting images to a USB device. This was to ensure best practices in preserving any metadata associated to the image files.
VI. Image files were then examined using JPEGsnoop and Pic2map. EXIF data was extracted and organized in Microsoft Excel.
Analysis to Determine Elevation Error
Elevation error at each location was determined in two different ways. Test series #1, elevation was determined by the known elevation at the ‘actual location’ that was acquired from SARTOPO. This actual location elevation was compared to the elevation being displayed by each test images EXIF data. With Test series #11, elevation was determined by the known elevation listed in the NGS survey marker data sheets. The elevation listed in the data sheets were compared to the elevation being displayed by each test image’s EXIF data.
With having a known and test data values, we can determine the percent error associated to each image by using the percent error formula listed in Figure 5. This formula was conducted in each test data set and a percent error was addressed for each test image file. After each test image file had a percent error associated to it, an average percent error was gathered for each device. These values are displayed in each test location data set.
12


Percentage Error =
x 100%
Va - G.
vA = approximate (measured) value vE = exact value
Figure 5: Percent Error Formula
Analysis to Determine Distance from Actual Location
In order to calculate an error between GPS coordinates, the distance between two points was the best way to display this type of error. From having an actual location coordinate and coordinate from each photo we can determine the length in meters between both coordinates.
This calculation was done using a Matlab script. This script is associated to the distance formula and Matlab defines the script as computing the lengths of the great circle arcs connecting pairs of points on the surface of a sphere, in each case the shorter arc is assumed ( https://www.mathwork s.com/help/map/ref/ distance.html#dll7e2 03211.
Matlab Script: [arclen, az] = distance(lat 1,Ion l,lat2,lon2)* 1000
The above script produced the number of meters from the actual location against the coordinates extracted from the experimental images. In each test, Tati’ and Ton2’ were the same values and represented the actual location point. They were measured against Tat2’ and Ton2’, which represented the latitude and longitude coordinates produced by each test image.
In each test data set, all coordinates gathered were converted to decimal degrees GPS format. This degree format does not change the location of the original acquired coordinate. This conversion was necessary for Matlab to be able to process the distance difference.
The distance from actual location from each image file and average actual location for each device test is displayed in each test data set in Chapter III.
13


III. TESTING LOCATIONS AND RESULTS FROM TEST #1 & #11
All tests will display a raw data table pertaining to the image file results and lists columns by file name, the device, the GPS satellite that device uses, date, time displayed in GMT-07:00,
GPS latitude reference, latitude, GPS longitude reference and elevation. All tests latitude and longitude coordinates are redacted by only the initial degrees portion of the coordinate.
Color coordination is also implemented to show a constant color for each device being tested. This color coordination stays consistent with the image file pertinent to the device used to show a source of where information originated from for the Garmin device photos.
For each test another table will follow that focuses on using the raw data to display what was analyzed. The analyzed data are in relation to the elevation percent error, average percent error of elevation, the distance each image file from the actual point (meters) and the average distance from the actual point (meters) relating to the device.
Test #1 - Rural Location #1
Rural Location #1, used the iPhone 6s, iPhone 7 Plus, Samsung SM-N960U, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 1. Below will list some observations listed in these data sets.
14


Table 1: Test I Rural Location #1 Coordinates and Elevation
File Name Device GPS Sat Date Time(GMT- 07:00) GPS Lat Latitude GPS Long Longitude Elevation Actual Location N 32'22" W °14'40" 227.99
53944444444 0.244444444
IMG_0118.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:04 N 34' 24.80” W 19' 26.06" 87
IMG_0119_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:05 N 34' 24.80" W 19' 26.06” 87
IMG_0120_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:06 N 32'21.81” W 14' 40.53” 232
IMG_0121.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:07 N 32'21.81" W 14' 40.53" 232
IMG_0122_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:14 N 32' 21.93" W 14' 40.72" 226
IMG_0123.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:14 N 32' 21.93" w 14' 40.72" 226
IMG_0124_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:17 N 32' 21.93" w 14' 40.72" 226
IMG_0125.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:03:18 N 32' 21.93" w 14' 40.72” 226
IMG_1934.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:02:04 N 32' 21.98" w 14' 40.58” 222
IMG_1935.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:02:12 N 32'21.98" w 14’ 40.50" 222
IMG_1936.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:02:33 N 32' 22.03" w 14' 40.61" 221
IMG_1937.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:02:38 N 32' 22.04" w 14’ 40.64" 222
20190916_100147.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:01:47 N 32' 22 02" w 14' 40.58" 207
20190916 100218.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:02:18 N 32' 22.03" w 14' 40.58” 206
20190916_100248_Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:02:48 N 32' 22.04" w 14' 40.56” 204
20190916_100257_Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:02:57 N 32' 22.03" w 14' 40.58” 206
IMG_0118.JPG Garmin Etrex 20x Garmin Default 9/16/2019 10:03:04 N 32.367' w 14.677' 222.504
IMG 1935.JPG Garmin Etrex 20x Garmin Default 9/16/2019 10:02:12 N 32.366' w 14.677' 222.199
IMG_0120_Airplane.JPG Garmin GPSmap 62s Garmin Default 9/16/2019 10:03:06 N 32'22.7” w 14'36.4” 223.723
IMG_1934.JPG Garmin GPSmap 62s Garmin Default 9/16/2019 10:02:04 N 32'22.7” w 14'36.5" 224.333
Table 2 reveals that the iPhone 6s, TMG_0119_Airplane.jpg’ and TMG_0120_Airplane .jpg’ had a change of distance from 7256.8 meters to 12.9379 meters from the actual location. By referencing Figure 6 the difference gap is shown. This is mainly due to the device trying to update its location to a much more precise one. This is worth noting because the device was in airplane mode and possibly still trying to update its own location.
Figure 6: Rural Location #1 - Anomaly ‘IMG 0119 Airplane.jpg ’ Location Compared to Actual
Location (SARTOPO Map)
15


Consistent distance of approximate 12 meters - 15 meters away from the actual location are acquired from the iPhone 6s, iPhone 7 Plus, Samsung SM-N960U and the Garmin eTrex 20x. Garmin GPSmap 62s displays a location roughly 80 meters away and stays consistent with producing that result. Figure 7 and 8, display SARTOPO and Google Earth maps displaying an overall view of all image file GPS coordinates plotted compared to the actual location.
Figure 7: Rural Location #1 -Image GPS Coordinates vs Actual Location (SARTOPO map)
Figure 8: Rural Location #1-Image GPS Coordinates us Actual Location (Google Earth map)
16


As for elevation, the percent error across all devices is consistent pertaining to each device. Anomalies with the iPhone 6s were displayed initially with TMG_0118.jpg’ and ‘IMG 0119_Airplane.jpg’ showing an initial percent error of 61.81%. This is then adjusted to a .87% error from TMG_0122_Airplane.jpg’ and stays consistent with the three image files that followed. It is worth noting that the Samsung SM-N960U’s elevation error was the worse with an approximate elevation percent error of 9% - 10%.
Table 2: Rural Location #1 - Percent Error Elevations and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG_0118.JPG iPhone 6s 61.84% 7256.8
IMG_0119_Airplane.JPG iPhone 6s 61.84% 7256.8
IMG_0120_Airplane.JPG iPhone 6s 1.76% 12.9379
IMG_0121.JPG iPhone 6s 1.76% 12.9379
IMG_0122_Airplane.JPG iPhone 6s 0.87% 15.7345
IMG_0123.JPG iPhone 6s 0.87% 15.7345
IMG_0124_Airplane.JPG iPhone 6s 0.87% 15.7345
IMG_0125.JPG iPhone 6s 0.87% 16.34% 15.7345 1825.301725
IMG_1934.JPG iPhone 7 Plus 2.63% 12.5108
IMG_1935.JPG iPhone 7 Plus 2.63% 10.9602
IMG_1936.JPG iPhone 7 Plus 3.07% 13.2867
IMG_1937.JPG iPhone 7 Plus 2.63% 1.80% 14.063 12.705175
20190916_100147.jpg Samsung SM-N960U 9.21% 12.5108
20190916_100218.jpg Samsung SM-N960U 9.65% 12.5108
20190916_100248_Airplane.jpg Samsung SM-N960U 10.52% 12.5108
20190916_100257_Airplane.jpg Samsung SM-N960U 9.65% 6.25% 12.5108 12.5108
IMG_0118.JPG Garmin Etrex 20x 2.41% 13.2867
IMG_1935.JPG Garmin Etrex 20x 2.54% 6.21% 13.2867 13.2867
IMG_0120_Airplane.JPG Garmin GPSmap 62s 1.87% 80.6975
IMG_1934.JPG Garmin GPSmap 62s 1.60% 5.93% 78.4449 79.5712
Test #1 - Rural Location #11
Rural Location #11 used the iPhone 6s, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. It is worth noting that a Samsung device was not utilized in this test. The raw data associated to this test is displayed in Table 3. Below will list some observations listed in these data sets.
17


Table 3: Rural Location #11 — Coordinates and Elevation
File Name Device GPS Sat Date Time (GMT -07:00) GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 32'23" wor 788.2128
32.384' 07.015'
IMG_0131.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:12:25 N 32' 22.81" W 7’ 1.68" 801
IMG_0132_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:12:31 N 32' 22.84" W 7' 1.79” 799
IMG_0134_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:12:58 N 32' 22.79" w 7' 1.74" 796
IMG_0135.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:13:05 N 32* 21.56" w 7' 0.89" 795
IMG_1950.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:09:29 N 32'41.60" w 23' 44.54" 5
IMG_1951_Airplane.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:09:42 N 32’ 23.43” w 7' 1.88" 794
IMG_1952_Airplane.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:11:52 N 32' 24.02” w 7' 1.68" 849
IMG_1953.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/17/2019 11:12:01 N 32' 23.28" w 7’ 1.35" 813
IMG_0131.JPG Garmin eTrex 20x Garmin Default 9/17/2019 11:12:25 N 32.372' w 7.028’ 793.6992
IMG_0132_Airplane.JPG Garmin eTrex 20x Garmin Default 9/17/2019 11:12:31 N 32.370' w 7.035’ 793.6992
IMG_1952.JPG Garmin eTrex 20x Garmin Default 9/17/2019 11:11:52 N 32.369' w 7.026’ 775.1064
IMG_1953.JPG Garmin eTrex 20x Garmin Default 9/17/2019 11:12:01 N 32.370' w 7.029' 790.6512
IMG_0134_Airplane.JPG GPSmap 62s Garmin Default 9/17/2019 11:12:58 N 32' 23.5” w 06' 57.5" 779.6784
IMG_0135.JPG GPSmap 62s Garmin Default 9/17/2019 11:13:05 N 32' 23.3” w 06' 57.5" 786.6888
IMG_1950.JPG GPSmap 62s Garmin Default 9/17/2019 11:09:29 N 32' 23.4” w 06' 56.8" 645.5664
IMG_1951.JPG GPSmap 62s Garmin Default 9/17/2019 11:09:42 N 32' 23.5” w 06' 56.4" 665.988
Table 4 and Figure 9 display an anomaly taking place. Starting with the iPhone 6s and looking at ‘IMG_0131.jpg’ and TMG_0135.jpg’. There is a change of distance from actual, starting from a close distance and exceeding to one that is more than double. Between these two image files, airplane mode is being turned on and producing similar coordinates to file ‘IMG_0131.jpg’. This change to airplane mode might be the cause for the sudden increase of distance possibly having the device relying on other connections.
Figure 9: Rural Location MI Image GPS Coordinates vs Actual Location (SARTOPO map)
18


Looking at the iPhone 7 Plus and at TMG_1950.jpg’, we have an initial distance of 21,717.1 meters away from the actual location. Within 13 seconds after that image file was taken, ‘IMG_1951_Airplane.jpg’ produced an image that was 22.9658 meters away from the actual location. This brings out another observation being made with the switch to airplane mode. Continuing with the same device the GPS coordinates from images produced after the previous image are displayed and narrowing in on the actual location where the device is present (Figure 11).
26.96
Miles
Figure 10: Rural Location II - Anomaly IMG 1950.jpg Distance from Actual Location
(SARTOPO map)
Figure 11: Rural Location II Anomaly #2 - iPhone 7 Plus Distance Corrections to Actual
Location (SARTOPO Map)
19


Garmin eTrex 20x was stable in respect to its distance data points. The Garmin GPSmap62s had distances displaying approximately 77 meters away from the actual location and possessed jumps to approximately 100 meters.
Elevation was consistent with all devices except for the iPhone 7 Plus and ‘IMG1950 .jpg’. The image file produced an elevation of 5 meters, which in turn produced a 99.37% error from the actual elevation. Again, 13 secs after that image file was taken, TMG_1951_Airpla ne.jpg’ produced an image that was 0.73% error from the actual elevation. Again, another trend with airplane mode change creating a more accurate result.
Table 4: Rural Location II — Percent Error Elevations and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG_0131.JPG iPhone 6s 1.62% 16.2322
IMG_0132_Airplane JPG iPhone 6s 1.37% 18.0136
IMG_0134_Airplane JPG iPhone 6s 0.99% 17.6638
IMG_0135 JPG iPhone 6s 0.86% 1.21% 44.5393 24.112225
IMGJ950 JPG iPhone 7 Plus 99.37% 21717.1
IMG_1951_Airplane.JPG iPhone 7 Plus 0.73% 22.9658
IMG_1952_Airplane.JPG iPhone 7 Plus 7.71% 34.4722
IMG_1953.JPG iPhone 7 Plus 3.14% 27.74% 11.8232 5446.5903
IMG_0131 JPG Garmin eTrex 20x 0.70% 25.7939
IMG_0132_Airplane JPG Garmin eTrex 20x 0.70% 34.3707
IMG_1952 JPG Garmin eTrex 20x 1.66% 29.4528
IMG_1953.JPG Garmin eTrex 20x 0.31% 0.84% 29.4267 29.761025
IMG_0134_Airplane.JPG GPSmap 62s 1.08% 77.1332
IMG_0135 JPG GPSmap 62s 0.19% 76.068
IMGJ950 JPG GPSmap 62s 18.10% 91.9398
IMGJ951 JPG GPSmap 62s 15.51% 8.72% 100.8977 86.509675
Test #1 - Rural Location #111
Rural Location #111 used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 5. Below will list some observations listed in these data sets.
20


Table 5: Rural Location III - Coordinates and Elevation
File Name Device GPS Sat Date Time (GMT -07:00) GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 32'23" 07'or 788.213
32.384 07.015’
IMG_0150.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:47:58 N 32'21.49" W 7' 0.06" 701
IMG 0151 .JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:48:03 N 32' 22.08" W 7' 0.34" 737
20190920_105509.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/20/2019 10:55:09 N 32' 23.00" w 7' 1.00" 761
20190920_105513.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/20/2019 10:55:13 N 32' 23.00" w 7' 1.00" 761
IMG_1974.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:47:36 N 32'21.64" w 7' 1.77" 791
IMG_1975.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:47:42 N 32'21.90" w 7' 1.82" 790
IMG_1976.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:48:56 N 32' 22.03" w 7' 1.55" 791
IMG_1977.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:49:16 N 32' 22.45" w 7' 1.52" 790
IMG_1978.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 10:49:30 N 32' 22.49" w 7' 1.98" 790
20190920_105513.jpg Garmin eTrex 20x Garmin Default 9/20/2019 10:55:09 N 32.376' w 7.025' 809.853
IMG_0150.JPG Garmin eTrex 20x Garmin Default + Glonass 9/20/2019 10:47:58 N 32.375' w 7.018' 802.843
IMG_1974.JPG Garmin eTrex 20x Garmin Default + Glonass 9/20/2019 10:47:36 N 32.376' w 7.021' 801.624
IMG 1978.JPG Garmin eTrex 20x Garmin Default 9/20/2019 10:49:30 N 32.381' w 07.020' 811.682
IMG_0151.JPG GPSmap 62s Garmin Default 9/20/2019 10:48:03 N 32' 23.0" w 6' 57.0" 799.795
20190920_105509.jpg GPSmap 62s Garmin Default 9/20/2019 10:55:09 N 32' 23.1" w 6' 57.1" 818.083
IMG_1977.JPG GPSmap 62s WAAS/EGNOS 9/20/2019 10:49:16 N 32' 22.9" w 6' 56.8" 812.292
IMG_1975.JPG GPSmap 62s Garmin Default 9/20/2019 10:47:42 N 32' 22.9" w 6' 56.8" 802.234
From referencing Table 6, the iPhone 6s had the worse cellular device average distance from actual location.
The Samsung Galaxy S6 had no error in relation to distance. This device was administered strictly on airplane mode and did not have cell service active. With this stipulation the Samsung Galaxy S6 took roughly 30 secs to 5 mins to acquire a GPS coordinate. This eludes to the device being able to produce a more than accurate coordinate. Mainly considering the coordinate being produced is being solely reliant on the GPS chip on the device.
The iPhone 7 Plus and Garmin eTrex 20x had variations between its distance, but nothing too alarming. Garmin GPSmap 62s again produced far distances from actual location results staying within approximately 80 meters - 91 meters.
Figure 12 displays a SARTOPO map of the image GPS coordinates compared to actual location.
21


Figure 12: Rural Location Mil Image GPS Coordinates vs Actual Location (SARTOPO Map)
Elevation showed the iPhone 6s had the worse average percent error and this might share
a correlation with that phone also producing poor distance results within this test. The other devices produced elevation errors that were not worth addressing.
Table 6: Rural Location Mil - Percent Error Elevation and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (ml


IMG_0150.JPG iPhone 6s 11.06% 50.9029
IMG_0151.JPG iPhone 6s 6.50% 8.78% 32.1303 41.5166
20190920_105509.jpg Samsung Galaxy S6 3.45% 0
20190920_105513.jpg Samsung Galaxy S6 3.45% 3.45% 0 0
IMG_1974.JPG iPhone 7 Plus 0.35% 45.309
IMG_1975.JPG iPhone 7 Plus 0.23% 38.847
IMG_1976.JPG iPhone 7 Plus 0.35% 32.2155
IMG_1977.JPG iPhone 7 Plus 0.23% 20.866
IMG_1978.JPG iPhone 7 Plus 0.23% 0.28% 26.1635 32.6802
20190920_105513.jpg Garmin eTrex 20x 2.75% 17.3806
IMG_0150.JPG Garmin eTrex 20x 1.86% 15.645
IMG_1974.JPG Garmin eTrex 20x 1.70% 14.4142
IMG 1978.JPG Garmin eTrex 20x 2.98% 2.32% 6.4513 13.472775
IMG_0151.JPG GPSmap 62s 1.47% 86.4496
20190920_105509.jpg GPSmap 62s 3.79% 84.1793
IMG_1977.JPG GPSmap 62s 3.05% 91.1837
IMG_1975.JPG GPSmap 62s 1.78% 2.52% 91.1837 88.249075
22


Test #1 - Rural Location #IV
Rural Location #IV used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 7. Below will list some observations listed in these data sets.
Table 7: Rural Location MV Rural MV Coordinates and Elevation
File Name Device GPS Sat Date Time (GMT -07:00) Location GPS Lat Latitude GPS Lon g Longitude Elevation
Actual Location 32'22" 14'40" 227.99
32.367' 14.669'
IMG_0152.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 11:16:42 Vista N 32'21.87” W 14' 40.75" 217
IMG_0153.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 11:16:46 Vista N 32'21.76” W 14' 40.53” 223
20190920_111700.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/20/2019 11:17:00 Vista N 32' 22.00" W 14' 40.00” 193
20190920_111705.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/20/2019 11:17:05 Vista N 32' 22.00” W 14' 40.00" 193
IMG_1979.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/20/2019 11:16:22 Vista N 32' 22.36" W 14' 39.70” 793
IMG_1980.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/21/2019 11:16:27 Vista N 32'21.72” W 14' 40.50" 196
IMG_1981.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/22/2019 11:16:31 Vista N 32' 21.82” w 14' 40.33” 222
IMG_1982.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/23/2019 11:17:46 Vista N 32'21.08” w 14' 39.81" 223
IMG_1983.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/24/2019 11:18:15 Vista N 32'21.79” w 14' 40.47" 222
IMG_1984.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/25/2019 11:18:18 Vista N 32’21.83” w 14' 40.39” 219
IMG_0153.JPG Garmin eTrex 20x Garmin Default 9/26/2019 11:16:46 Vista N 32.365' w 14.675' 224.63
IMG_1981.JPG Garmin eTrex 20x Garmin Default 9/27/2019 11:16:31 Vista N 32.365' w 14.675' 223.418
IMG_1982.JPG Garmin eTrex 20x Garmin Default + GLONASS 9/28/2019 11:17:46 Vista N 32.366' w 14.674' 220.675
IMG_0152.JPG GPSmap 62s Garmin Default 9/29/2019 11:16:42 Vista N 32' 22.6" w 14' 36.4” 222.8
IMG_1980.JPG GPSmap 62s Garmin Default 9/30/2019 11:16:27 Vista N 32' 22.5” w 14' 36.4” 222.809
IMG_1984.JPG GPSmap 62s Garmin Default 10/1/2019 11:18:18 Vista N 32' 22.5" w 14' 36.4" 229.21
Referencing Table 8 and Figure 13, it is shown that the iPhone 7 Plus displays the worse average distance from actual location for a cellular device. The Samsung Galaxy S6 again displayed no error regarding distance. The iPhone 6s and the Garmin eTrex 20x produced a semi constant result. GPSmap 62s again stayed within its approximately 80 meter distance error.
Figure 13: Rural Location MV - Image GPS Coordinates vs Actual Location (SARTOPO Map)
23


Figure 14: Rural Location #IV- Image GPS Coordinates vs Actual Location (Google Earth
Map)
Figure 14 displays a Google Earth map to display a visual of the type of environment where this test was conducted. From data in Table 8, the iPhone 7 Plus had the worse average percent error. This again eludes to the elevation and distance error sharing the same error correlation.
However, what goes against this correlation is the Samsung Galaxy S6. The Galaxy S6 produced no error whatsoever with distance but displayed a 15.35% error in elevation. This further can elude to this type of trend being noticed based on the devices itself, rather than the distance and elevation error sharing the same type of error rate.
24


Table 8: Rural Location #IV - Percent Error Elevation and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG_0152.JPG iPhone 6s 4.82% 16.9494
IMG_0153.JPG iPhone 6s 2.19% 3.50% 14.038 15.4937
20190920J11700 jpg Samsung Galaxy S6 15.35% 0
20190920_111705.jpg Samsung Galaxy S6 15.35% 15.35% 0 0
IMG_1979.JPG iPhone 7 Plus 247.82% 12.7461
IMG_1980.JPG iPhone 7 Plus 14.03% 14.072
IMG_1981.JPG iPhone 7 Plus -2.63% 29.1718
IMG_1982.JPG iPhone 7 Plus 2.19% 29.1718
IMG_1983.JPG iPhone 7 Plus 2.63% 12.1253
IMG_1984.JPG iPhone 7 Plus 3.94% 44.66% 10.2131 17.91668333
IMG_0153.JPG Garmin eTrex 20x 1.47% 11.2254
IMG_1981.JPG Garmin eTrex 20x 2.01% 11.2254
IMG_1982 JPG Garmin eTrex 20x 3.21% 2.23% 9.3783 10.6097
IMG_0152.JPG GPSmap 62s 2.28% 79.4234
IMG_1980.JPG GPSmap 62s 2.27% 79.4
IMG_1984.JPG GPSmap 62s 0.54% 1.69% 79.4 79.4078
Test #1 - Suburb Location #1
Suburb Location #1 was used the iPhone 6s, Samsung SM-N960U, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 9. Below will list some observations listed in these data sets.
Table 9: Suburb Location #1 - Coordinates and Elevation
File Name Device GPS Sat Date Time(GMT -07:00) GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 31'50" 26'20" 60.96
31.836'. 26.332'
IMG 0126.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:29:54 N 31’ 50.71” W 26' 19.06" 58
IMG_0127.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:29:58 N 31’51.09" W 26' 19.81" 65
IMG_0128_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:30:05 N 31’ 50.47" w 26' 19.70" 63
IMG_0130_Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:30:09 N 31’50.47" w 26' 19.70" 62
IMG_1938.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:28:35 N 31’50.48" w 26' 19.70" 60
IMG 1939.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:29:05 N 31’50.23" w 26' 19.72" 60
IMG_1940.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:29:15 N 3V 50.12" w 26' 19.94" 60
IMG_1941.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 10:29:44 N 31' 50.28" w 26' 20.08" 60
20190916 102840.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:28:40 N 31' 50.57” w 26' 19.76" 35
20190916_102850.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:28:50 N 31'50.57" w 26' 19.76" 35
20190916_102906_Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:29:06 N 31' 50.57" w 26’ 19.76" 35
20190916_102914_Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 10:29:14 N 32- 22.04” w 14' 40.56" 204
IMG_0126.JPG Garmin GPSmap 62s Garmin Default 9/16/2019 10:29:54 N 31’50.8" w 26' 15.7" 60.6552
IMG_1940.JPG Garmin GPSmap 62s Garmin Default 9/16/2019 10:29:15 N 31'50.8" w 26' 15.6" 61.8744
IMG_0127.JPG Gamin eTrex 20x Garmin Default 9/16/2019 10:29:58 N 31.837' w 26.332' 58.8264
IMG_0130_Airplane.JPG Gamin eTrex 20x Garmin Default 9/16/2019 10:30:09 N 31.836’ w 26.331' 58.2168
IMG_1938.JPG Gamin eTrex 20x Garmin Default 9/16/2019 10:28:35 N 31.837' w 26.331' 58.8264
In referencing Table 10, the biggest anomaly that stands out is image file ‘20190916 1 02914_ Airplane.jpg’ from the Samsung SM-N960U. This image file was 15,164.6 meters from the actual location and a 90.60% error rate regarding elevation. It is interesting that this image file was the last image taken during the test. The SM-N060U previously captured image files
25


producing a 18.3948 meter distance from actual and a 42.59% elevation error rate. The cellular device was placed in airplane mode. However, this could be something going on in the background in the device priority list of how it acquires location and was trying to correct itself. Figure 15 displays the distance between the two points.
Figure 15: Suburb Location #1 - Anomaly Samsung Note 9 ‘20190916 102914 Airplane.JPG’
Distance from Actual Location (SARTOPO Map)
Another anomaly was with the iPhone 6s. This initially produced an approximate 30 meter distance from actual location, but then corrected to an approximate 15 meter from actual point. The other devices stayed consistent with the trends being associated to them in previous tests. Figures 16 and 17 displays the test image GPS coordinates vs actual location.
Figure 16: Suburb Location #1- Image GPS Coordinates vs Actual Location (SARTOPO Map)
26


Figure 17: Suburb Location #1 -Image GPS Coordinates vs Actual Location (Google Earth
Map)
Table 10: Suburb Location #1 - Percent Error and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG 0126.JPG iPhone 6s 4.86% 30.079
IMG 0127.JPG iPhone 6s 6.63% 33.5851
IMG_0128_Airplane.JPG iPhone 6s 3.35% 15.7413
IMG_0130_Airplane.JPG iPhone 6s 1.71% 4.13% 15.7413 23.786675
IMG 1938.JPG iPhone 7 Plus 1.57% 15.7413
IMG 1939.JPG iPhone 7 Plus 1.57% 8.6164
IMG_1940.JPG iPhone 7 Plus 1.57% 3.4256
IMG_1941.JPG iPhone 7 Plus 1.57% 1.57% 9.1974 9.245175
20190916_102840.jpg Samsung SM-N960U 42.59% 18.3948
20190916 102850.jpg Samsung SM-N960U 42.59% 18.3948
20190916_102906_Airplane.jpg Samsung SM-N960U 42.59% 18.3948
20190916_102914_Airplane.jpg Samsung SM-N960U 234.65% 90.60% 15164.6 3804.9461
IMG 0126.JPG Garmin GPSmap 62s 0.50% 95.8688
IMG_1940.JPG Garmin GPSmap 62s 1.50% 1.00% 98.1301 96.99945
IMG_0127.JPG Gamin eTrex 20x 3.50% 6.8512
IMG_0130_Airplane.JPG Gamin eT rex 20x 4.50% 6.0309
IMG_1938.JPG Gamin eTrex 20x 3.50% 3.83% 7.7019 6.861333333
Test #1 - Suburb Location #11
Suburb Location #11 used the Samsung SM-N960U, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 11. Below will list some observations listed in these data sets.
27


Table 11: Suburb Location #11- Coordinates and Elevation
File Name Device GPS Sat Date Time(GMT 07:00) GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 31'50" 26'20" 60.96
31.836' 26.332'
IMG_1943.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 16:46:22 N 31'50.22" W 26' 20.03" 60
IMG_1944.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 16:46:26 N 31'50.40" W 26' 20.08" 60
IMG 1945JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/16/2019 16:46:43 N 31'50.18” W 26’ 19.86" 60
20190916_164615.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 16:46:15 N 31' 50.45” W 26’ 20.02" 36
20190916_164635 Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 16:46:35 N 31'50.45" W 26' 20.02” 36
20190916_164659.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 16:46:59 N 31' 50.45" W 26' 20.02” 36
20190916 164714Airplane.jpg Samsung SM-N960U A-GPS, GLONASS, GALILEO, BDS 9/16/2019 16:47:14 N 31'51.34" W 26' 19.55" 36
IMG_1944.JPG GPSmap 62S Garmin Default 9/16/2019 16:46:26 N 31'50.2" W 26' 16” 78.0288
IMG_1945.JPG GPSmap 62S Garmin Default 9/16/2019 16:46:43 N 31'50.3" W 26' 16" 72.5424
20190916 164615-jpg GPSmap 62S Garmin Default 9/16/2019 16:46:15 N 31'50.3" W 26’ 16.1” 87.1728
20190916_164659.jpg Gamin eTrex 20x Garmin Default 9/16/2019 16:46:59 N 31.835' W 26.332 62.1792
In reference to Table 12, the iPhone 7 Plus had a slight fluctuation from close distances to actual location to farther ones. This is not so alarming seeing this type of movement due to the device constantly trying to adjust its location to provide a precise location.
It is worth noting that the Samsung device provided the worse average distance from actual location and the worse elevation percent error compared to the iPhone 7 Plus. The Garmin eTrex provided the best average distance from actual point.
Figures 18 and 119 display SARTOPO and Google Earth maps displaying distances between each device image file coordinates.
Figure 18: Suburb Location #11-Image GPS Coordinates vs Actual Location (SARTOPO Map)
28


Figure 19: Suburb Location #11-Image GPS Coordinates vs Actual Location (Google Earth
Map)
With elevation the iPhone 7 Plus provided the best percent error of 1.57% compared to all devices. The Samsung device and Garmin units surprising provided relatively high percent errors compared to the iPhone 7 Plus, ranging from 22.50% to 40.94%.
Table 12: Test #VI Suburb #11 - Percent Error Elevation and Average Distance from Actual
File Nam* Device % Error Elevation Average % Error Elevation Distance From *v#rafl# Actual(m) Distance from 1 ' Actual (m)


IMG, 1943 JPG iPhone 7 Plus 1.57% 6.717
IMG, 1944 JPG iPhone 7 Plus 1.57% 12.4527
IMG 1945 JPG iPhone 7 Plus 1.57% 1.57% 6.0309 8.4002
20190916,164615 jpg Samsung SM -N960U 40.94% 14.4763
20190916 164635 Airplane jpg Samsung SM-N960U 40.94% 14.4763
20190916,164659jpg Samsung SM-N960U 40.94% 14.4763
20190916 164714 Airplane ipg Samsung SM-N960U 40.94% 40.94% 42.1906 21.404875
IMGJ944 JPG GPSmap 62S 28.00% 86.7207
IMGJ945 JPG GPSmap 62S 19.00% 86.9201
20190916,164615 jpg GPSmap 62S 43.00% 30.00% 84.2085 85.94976667
20190916,164659 jpg Gamin eTrex 20x 2.00% 22.SO% 3.6817 3.6817
Test #1 - Suburb Location #111
Suburb Location #111 used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin
eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table
13. Below will list some observations listed in these data sets.
29


Table 13: Suburb Location III Coordinates and Elevation
File Name Device GPS Sat Date Time GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 31'50" °26'20" 60.96
31.836 °26.333'
IMG_0138.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:16 N 31'50.41" W 26' 20.27" 36
IMG_0139.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:22 N 31'50.19” W 26' 19.70" 59
IMG_0141.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:50 N 31'50.12" w 26’ 19.83" 60
IMG_0140JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:50 N 31'50.12" w 26' 19.83" 60
IMG_0142JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:51:00 N 31'49.81" w 26' 19.45” 61
IMG_0143.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:52:32 N 31'50.08" w 26' 19.67" 62
IMG_0144.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:52:32 N 31'50.08” w 26' 19.67" 62
IMG_1955.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:49:52 N 31' 51.29" w 26' 20.36" 61
IMG_1956.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:00 N 31'50.48" w 26' 20.11" 60
IMG_1957.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:50:01 N 31' 50.48" w 26' 20.11" 60
IMG_1958.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:51:08 N 31'50.08" w 26' 20.30" 60
IMG_1959.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:51:15 N 31'50.03” w 26' 19.75" 60
IMG_1960.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:52:21 N 31' 50.19" w 26' 19.89" 56
IMG 1961 JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 10:52:21 N 31' 50.19" w 26' 19.89" 56
20190919 110559.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/19/2019 11:05:59 N 31'50.00” w 26' 20.00" 38
IMG_0141.JPG Garmin eTrex 20x Gamin Default 9/19/2019 10:50:50 N 31.837' w 26.332' 51.2064
IMG_0142.JPG Garmin eTrex 20x Gamin Default 9/19/2019 10:51:00 N 31.836* w 26.332’ 56.388
IMG_0143.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 10:52:32 N 31.836' w 26.331' 59.436
IMG_1958.JPG Garmin eTrex 20x Gamin Default 9/19/2019 10:51:08 N 31.837' w 26.333' 51.816
IMG_1961.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 10:52:21 N 31.837' w 26 330' 60.0456
IMG_0138.JPG GPSmap 62s Gamin Default 9/19/2019 10:50:16 N 31'50.8" w 26' 15.7” 58.2168
IMG_1955.JPG GPSmap 62s Gamin Default 9/19/2019 10:49:52 N 31' 50.5" w 26' 15.1" 59.7408
IMG_1956.JPG GPSmap 62s Gamin Default 9/19/2019 10:50:00 N 31' 50.5" w 26' 15.5" 70.7136
Table 14 shows the Samsung Galaxy S6 possessed no error rate regarding distance but displayed a 37.66% error rate regarding elevation.
The iPhone 6s and iPhone 7 Plus average distances from actual location did not produce alarming results. They were consistent and close to the actual location with 8 meter to 13 meter differences.
Garmin eTrex 20x had a better average distance from actual location than the iPhone 7 Plus and iPhone 6s. Garmin GPSmap 62s still produced poor results by have the worse average distance from actual location.
Figure 20 displays the SARTOPO map of all image file GPS coordinates compared to actual location.
30


Figure 20: Suburb Location Mil Image GPS Coordinates vs Actual Location (SARTOPO Map)
With elevation, the iPhone 6s and iPhone 7 Plus had a different elevation profile that was not too similar. The iPhone 6s produced an average elevation percent error of 7.26% and the iPhone 7 Plus produced an average percent error of 3.23%. From these devices being only a year device generation apart, it was interesting to see that these error rates will not be relatively closer.
The Garmin eTrex 20x did not produce a good elevation profile in comparison to the cellular devices. Garmin GPSmap 62s again produced poor results by having the worse average elevation percent error.
Table 14: Suburb Location Mil - Percent Error Elevation and Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG 0138.JPG IPhone 6s 40.94% 13.7274
IMG_0139.JPG IPhone 6s 3.22% 8.3513
IMG 0141.JPG IPhone 6s 1.57% 4.5647
IMG_0140JPG iPhone 6s 1.57% 4.5647
IMG_0142JPG iPhone 6s 0.07% 12.9396
IMG 0143 JPG iPhone 6s 1.71% 7.3549
IMG 0144 JPG iPhone 6s 1.71% 7.26% 7.3549 8.408214286
IMG 1955.JPG iPhone 7 Plus 0.07% 40.781
IMG_1956 JPG IPhone 7 Plus 1.57% 14.643
IMG_1957.JPG iPhone 7 Plus 1.57% 14.643
IMG 1958 JPG iPhone 7 Plus 1.57% 7.3549
IMGJ959 JPG iPhone 7 Plus 1.57% 5.5649
IMG 1960 JPG iPhone 7 Plus 8.14% 6.0309
IMG 1961.JPG iPhone 7 Plus 8.14% 3.23% 6.0309 13.57837143
20190919 110559 jpg Samsung Galaxy S6 37.66% 37.66% 0 0
IMG_0141 JPG Garmin eTrex 20x 16.00% 6.8512
IMG_0142 JPG Garmin eTrex 20x 48.39% 5.7739
IMG 0143.JPG Garmin eTrex 20x 56.41% 6.0309
IMG_1958 JPG Garmin eTrex 20x 36.36% 6.6717
IMG 1961 JPG Garmin eTrex 20x 58.01% 43.03% 7.7253 6.55045
IMG 0138 JPG GPSmap 62s 53.20% 95.8688
IMG_1955 JPG GPSmap 62s 57.21% 107.0751
IMG_1956.JPG GPSmap 62s 86.09% 65.50% 98.6055 100.5164667
31


Test #1 - Suburb Location #IV
Suburb Location #IV used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 15. Below will list some observations listed in these data sets.
Table 15: Suburb Location #IV- Coordinates and Elevation
File Name Device GPS Sat Date Time(GMT -07:00) GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 31'50" 26'20" 60.96
31.836' 26.332'
IMG_0145.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:18:33 N 31' 50.76” W 26' 19.42” 58
IMG_0146.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:18:33 N 31'50.76” W 26' 19.42” 58
IMG_0147.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:19:10 N 31'50.85” w 26' 19.59” 58
IMG_1962.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:18:42 N 31'50.76" w 26' 19.67” 58
IMG_1963.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:18:42 N 31’50.76" w 26' 19.67” 58
IMG_1965.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:18:48 N 31'50.58” w 26' 20.08” 56
IMG_1966.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 16:19:22 N 31'50.23” w 26' 20.00” 59
20190919_161930.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/19/2019 16:19:30 N 31' 49.00" w 26' 19.00” 40
20190919_161940.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/19/2019 16:19:40 N 31'49.00” w 26' 19.00” 40
IMG_1962.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 16:18:42 N 31.836' w 26.333’ 54 5592
IMG_1963.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 16:18:42 N 31.836' w 26.332' 54.5592
IMG_0145.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 16:18:33 N 31.838' w 26.333' 51.816
20190919_161940.jpg Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 16:19:40 N 31.832' w 26.330' 57.3024
20190919_161940.jpg GPSmap 62s Gamin Default 9/19/2019 16:19:40 N 31' 50.8" w 26' 15.7” 57.3024
IMG_0147.JPG GPSmap 62s Gamin Default 9/19/2019 16:19:10 N 31'50.9” w 26.333' 7" 59.436
Table 16 displayed the Samsung Galaxy S6 producing the worse average distance from actual location within cellular devices and the worse average elevation percent error categories. The two image files that were produced by the Galaxy S6 were also taken with 10 seconds apart and did not produce different results.
The iPhone 7 Plus outperformed all devices in distance and the GPSmap 62s outperformed all devices in elevation. Figure 21 displays an overall view of all image GPS coordinates vs the actual location.
32


Figure 21: Suburb Location #IV- GPS Coordinates vs Actual Location (SARTOPO Map)
There was one distinct change in distance with one device showing the device correcting itself closer to the actual location. The iPhone 7 Plus ‘IMG_1966.jpg’ file produced a distance of 6.717 meters from actual location. Previous images associated to this device produced an approximate 17 meter - 24 meter distance from actual location. The amount of time the device took for this correction was 40 seconds.
Table 16: Suburb Location #IV- Percent Error and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG_0145 JPG iPhone 6s 4.86% 26.8444
IMG_0146 JPG iPhone 6s 4.86% 26.8444
IMG_0147.JPG iPhone 6s 4.86% 4.86% 28.2765 ’ 27.32176667
IMG_1962 JPG iPhone 7 Plus 4.86% 24.6159
IMG_1963 JPG iPhone 7 Plus 4.86% 24.6159
IMG_1965 JPG iPhone 7 Plus 8.14% 17.8593
IMG_1966 JPG iPhone 7 Plus 3.22% 5.27% 6.717 18.452025
20190919_161930.jpg Samsung Galaxy S6 34.38% 38.0141
20190919_161940.jpg Samsung Galaxy S6 34.38% 34.38% 38.0141 38.0141
IMG_1962 JPG Garmin eTrex 20x 10.50% 5.614
IMG_1963 JPG Gannin eTrex 20x 10.50% 6.0309
IMG_0145.JPG Garmin eTrex 20x 15.00% 8.9296
20190919_161940 jpg Garmin eTrex 20x 6.00% 10.50% 5.1758 6.437575
20190919_161940 jpg GPSmap 62s 6.00% 96.222
IMG_0147.JPG GPSmap 62s 2.50% 2.50% 152.886 124.554
33


Test #1 - Urban Location #1
Urban Location #1 used the iPhone 7, iPhone 8, iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 17. Below will list some observations listed in these data sets.
Table 17: Urban Location #1 - Coordinates and Elevation
File Name Device GPS Sat Date Time GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 3r26" 40'49" 13.1064
31.426' 40.814'
IMG_2722.JPG iPhone 7 A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:58:39 N 31’ 23.43" W 40' 45.36” 45
IMG_2723.JPG iPhone 7 A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:58:51 N 31'25.01" W 40' 48.03" 16
IMG_0691.JPG iPhone 8 A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:58:03 N 31’25.97" W 40' 49.57" 13
IMG_0692.JPG iPhone 8 A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:58:14 N 31'25.61" W 40' 48.96" 14
IMG_0148.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:57:06 N 31'25.71" W 40' 48.96" 14
IMG_0149.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:57:23 N 31'25.25" W 40' 48.83" 15
IMG_1967.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:56:47 N 31'25.56" W 40' 48.88" 29
IMG_1968.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 9/19/2019 18:57:31 N 31' 20.74" W 40' 42.95" 12
20190919_190059.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/19/2019 19:00:59 N 31'26.00” W 40' 49.00" 0
20190919_190106.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 9/19/2019 19:01:06 N 31'26.00" W 40' 49.00" 0
IMG_0149.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 18:57:23 N 31.424' W 40.815’ 27.1272
IMG_0691.JPG Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 18:58:03 N 31.425' W 40.816' 28.3464
20190919_190106.jpg Garmin eTrex 20x Gamin Default + GLONASS 9/19/2019 19:01:06 N 31.432' W 40.819' 13.716
IMG_2723.JPG GPSmap62s Gamin Default 9/19/2019 18:58:51 N 31’25.8" W 40' 44.0" 65.2272
IMG_0692.JPG GPSmap62s Gamin Default 9/19/2019 18:58:14 N 31' 25.6" W 40' 43.9" 64.008
IMG_0148.JPG GPSmap62s Gamin Default 9/19/2019 18:57:06 N 31'25.8" W 40' 44.0" 60.0456
IMG_1967.JPG GPSmap62s Gamin Default 9/19/2019 18:56:47 N 31’25.7" W 40' 44.2" 56.9976
20190919_190059.jpg GPSmap62s Gamin Default 9/19/2019 19:00:59 N 31’ 25.3” W 40' 43.7” 57.6072
From looking at results from Table 18, iPhone 7 Plus had the worse results of average distance from actual location. By comparing TMG_1967.jpg’ to TMG_1968.jpg’ the initial image file produced a 13.5465 meter distance from actual location. Then an image taken 1 min and 16 seconds later produced a 208.5322 meter distance from actual location. This was the biggest jump seen in this test.
However, the iPhone 7 produced a similar type of circumstance comparing ‘IMG 2722. jpg’ and TMG_2723.jpg’. This circumstance was opposite from the iPhone 7 Plus, where the image file coordinate was first providing a far distance but then corrected to a closer distance.
34


It was of interest to see the iPhone 8, iPhone 6s, Samsung Galaxy s6 and the Garmin
eTrex 20x were able to provide a distance stable to one another.
Samsung Galaxy S6 was able to produce no error in distance again, which was surprising with being in an urban environment. However, its elevation percent error was 100%, in which the device was not able to produce an elevation whatsoever.
Figures 22 and 23 display SARTOPO and Google Earth maps displays image GPS coordinates compared to actual location.
Figure 22: Urban Location #1 -Image GPS Coordinates vs Actual Location (SARTOPO Map)
Figure 23: Urban Location #1 -Image GPS Coordinates vs Actual Location (Google Earth
Map)
35


Moving to overall elevation in this test, the Garmin GPSmap 62s produced the worse average elevation percent error rate at a 363.72%. The worse cellular device elevation percent error was the iPhone 7 at a 132.71%. The iPhone 8 was able to produce the best elevation average percent error profile at a 3.81%.
Table 18: Urban Location #1 - Percent Error Elevation and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG_2722.JPG iPhone 7 243.34% 111.4631
IMG_2723.JPG iPhone 7 22.08% 132.71% 36.6579 74.0605
IMG_0691 JPG iPhone 8 0.81% 12.4647
IMG_0692.JPG iPhone 8 6.82% 3.81% 11.1467 11.8057
IMG_0148.JPG iPhone 6s 6.82% 8.9296
IMG_0149.JPG iPhone 6s 14.45% 10.63% 22.4562 15.6929
IMG_1967.JPG iPhone 7 Plus 121.27% 13.5465
IMG_1968.JPG iPhone 7 Plus 8.44% 64.85% 208.5322 111.03935
20190919_190059.jpg Samsung Galaxy S6 100.00% 0
20190919_190106jpg Samsung Galaxy S6 100.00% 100.00% 0 0
IMG_0149.JPG Garmin eTrex 20x 106.98% 16.7519
IMG_0691.JPG Garmin eTrex 20x 116.28% 14.4763
20190919_190106.jpg Garmin eTrex 20x 4.65% 75.97% 3.8284 11.68553333
IMG_2723.JPG GPSmap62s 397.67% 108.4299
IMG_0692.JPG GPSmap62s 388.37% 110.5243
IMG_0148.JPG GPSmap62s 358.14% 107.8075
IMG_1967.JPG GPSmap62s 334.88% 103.9942
20190919_190059.jpg GPSmap62s 339.53% 363.72% 116.3724 109.42566
Test #11 - NGS Location #1 Urban #11
NGS Location #1 Urban #11, used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. This was the first test that utilized the NGS survey markers. The survey marker was in downtown Portland, Oregon, USA. Figure 24 displays a photo of the NGS survey maker with the NGS data sheet that displays the GPS coordinates and the elevation of the marker. This test was the initial implementation of putting our devices on airplane mode, powering off device, powering on device ensuring airplane mode is still active and capturing the initial photo in the test series. The raw data associated to this test is displayed in Table 19. Below will list some observations listed in these data sets.
36


The NGS Data Sheet
See file dsdata.pdf for more information about the datasheet.
PROGRAM = datasheet95, VERSION = 8.12.5.4 1 National Geodetic Survey, Retrieval Date = OCTOBER 16, 2019 AB7226 *********************************************************************** AB7226 DESIGNATION - 54 RESET
AB7226 PID - AB7226
AB7226 STATE/COUNTY- OR/MULTNOHAH AB7226 COUNTRY - US
AB7226 USGS QUAD - PORTLAND (1990)
AB7226
AB7226 'CURRENT SURVEY CONTROL
AB7226
AB7226* NAD 83(1986) POSITION- 45 31 08.1 (N) 122 40 41.8
AB7226* NAVD 88 ORTHO HEIGHT - 14.86 (*eters) 48.8
AB7226 ________________________________________________________
AB7226 GEOID HEIGHT - -22.813 (peters)
AB7226 VERT ORDER - THIRD
(N) HDHELD2 (feet) RESET
GEOID18
Figure 24: NGS Location #1 Urban #11 - NGS Data Sheet Rural Environment and Survey marker
54 RESET
Table 19: NGS Location #1 Urban #11 - Coordinates and Elevation
File Name Device GPS Sat Date Time GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 3T08.1" 40'41.8" 14.86
0.5189 0.6783
IMG_0160-Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 6:16:30 PM (Unknown) N W -
IMG_0161.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 18:18:19 (GMT-07:00) N 0.518883 W 0.678275 17
20191005_182353-Airplane.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 10/5/2019 18:23:53 (GMT-07:00) N 0.519167 w 0.678333 0
IMG_2013-Airplane.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 18:15:41 (Uknnown) N w
IMG_2014-Airplane.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 18:15:44 (Unknown) N - w -
IMG_2015.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 18:17:37 (GMT -07:00) N 0.518536 w 0.681397 111
IMG 2016.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/5/2019 18:17:52 (GMT -07:00) N 0.518975 w 0.678589 17
IMG_0160-Airplane.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/5/2019 6:16:30 PM (Unknown) N 31.150' w 40.697' 21.336
20191005_182353-Airplane.jpg Garmin eTrex 20x Garmin Default + GLONASS 10/5/2019 18:23:53 (GMT -07:00) N 31.145' w 40.701' 18.5928
IMG 2013-Airplane.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/5/2019 18:15:41 (Unknown) N 31.150' w 40.694' 14 3256
IMG_0161.JPG GPSmap 62s Garmin Default 10/5/2019 18:18:19 (GMT -07:00) N 31'09.7” w 40'37.0" 9.144
IMG_2015.JPG GPSmap 62s Garmin Default 10/5/2019 18:17:37 (GMT -07:00) N 31'09.7" w 40'37.0" 10.0584
IMG_2016.JPG GPSmap 62s Garmin Default 10/5/2019 18:17:52 (GMT -07:00) N 31'09.7” w 40'37.0" 11.2776
Table 20 shows the iPhone 7 Plus with both the worse average distance from actual location and average elevation percent error. This could be due to the fact of this device initially having problems acquiring a location. By looking at ‘IMG_2013-Aiplane.jpg’ and ‘IMG 2014-Airplane.jpg’ you can see that these initial photos did not obtain a GPS location or elevation.
This is due to the devices camera app location settings set to ‘while using the app’, which in turn takes approximately 25 secs for the device to acquire any GPS location. I believe this same result happened with the iPhone 6s ‘IMG_0160-Airplane.jpg’ file. However, its next image file was
37


able to provide a pretty accurate distance from actual location. Figure 25 displays the image GPS coordinates compared to the actual GIS survey marker location.
Figure 25: NGS Location #1 Urban II - Image GPS Coordinates from Actual Location
(SARTOPOMap)
It is worth noting that the Samsung Galaxy S6 was not able to provide an elevation profile. However, did provide a location point of 29.8002 meters from actual location. The stand-alone GPS units produced similar results as previous tests administered.
Table 20: NGS Location #1 Urban #11 - Percent Error and Average Distance from Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m] Average Distance from Actual (m)


IMG_0160-Airplane JPG iPhone 6s
IMG 0161.JPG iPhone 6s 14.40% 14.40% 2.7142 2.7142
20191005_182353-Airplane.jpg Samsung Galaxy S6 100.00% 100.00% 29.8002 29.8002
IMG_2013-Airplane.JPG iPhone 7 Plus
IMG_2014-Airplane.JPG iPhone 7 Plus
IMG_2015.JPG iPhone 7 Plus 646.97% 244.6635
IMG_2016.JPG iPhone 7 Plus 14.40% 330.69% 24.0112 134.33735
IMG_0160-Airplane JPG Garmin eTrex 20x 43.58% 29.673
20191005_182353-Ai rplane jpg Garmin eTrex 20x 25.12% 20.751
IMG_2013-Airplane JPG Garmin eTrex 20x 3.60% 24.10% 30.0349 26.81963333
IMG_0161.JPG GPSmap 62s 38.47% 117.7073
IMG_2015.JPG GPSmap 62s 32.31% 117.7073
IMG_2016.JPG GPSmap 62s 24.11% 31.63% 117.7073 117.7073
Test #11 - NGS Location #11 Rural #V
NGS Location #11 Rural #IV, used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus,
Garmin eTrex 20x and the Garmin GPSmap62s. This was the second test that utilized the NGS
38


survey markers. The survey marker was at a rural site used in a couple previous tests in our rural environment series outside of Portland, Oregon, USA. Worth noting, this test was a couple meters away from the previous tests’ actual location. Figure 26 displays a photo of the NGS survey maker with the NGS data sheet that displays the GPS coordinates and the elevation of the marker. This test also implemented our second test putting our devices on airplane mode first, powering off device, powering on device and capturing a photo. The raw data associated to this test is displayed in Table 21. Below will list some observations listed in these data sets.
3D2197- SAD 83(1991) POSITION- 45 32 21.5L9M(N) 122 14 46.72899(31) AOIUSTEO 1D2197* UVD 88 ORTHO HEIGHT • 225. (ttters) 738. (feet) SCALED 3D2197 _______________________________________________________________________________
iOGRAH = datastieet95, VERSION - 8.12.5.4
National Geodetic Survey, Retrieval Date = OCTOBER 16, 2819
3D2197....................................................................
3D2197 DESIGNATION • BN 1D2197 PID ID2197 STATE/COUNTY-ID2197 COUNTRY -
ID2197 USGS QUAD -
ID2197 ID2197
ID2197 ______________
RD2197
OR/HULTNOWH
US
BRIDAL VEIL (1994)
•CURRENT SURVEY CONTROL
GE0ID18
DEFLEC18
-22.392 (ueters) 5.83 (seconds)
- THIRD
Figure 26: NGS Location II Urban #V-NGS Data Sheet Urban Environment and Survey
Marker RD 2197
39


Table 21: NGS Location MI Rural MV - Coordinates and Elevation
File Name Device GPS Sat Date Time GPS Lat Latitude GPS Long Longitude Elevation
Actual Location 32*21.52" 14'40.72" 225
32.359’ 14.679’
IMG 0162-Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:54:54 (GMT -07:00) N 0.539264 W 0.244714 222
IMG 0163-Airplane.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:54:55 (GMT -07:00) N 0.539264 W 0.244714 222
IMG_0164.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:55:01 (GMT-07:00) N 0.539331 w 0.244683 221
IMG 0165.JPG iPhone 6s A-GPS, GLONASS. GALILEO. QZSS 10/6/2019 07:55:02 (GMT -07:00) N 0.539331 w 0.244683 221
IMG_0166.JPG iPhone 6s A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:55:17 (GMT -07:00) N 0.539344 w 0.244644 223
20191006_075656.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 10/6/2019 07:56:56 (GMT -07:00) N 0.539444 w 0.244722 224
20191006_075658.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 10/6/2019 07:56:57 (GMT -07:00) N 0.539444 w 0.244722 224
20191006_075659.jpg Samsung Galaxy S6 A-GPS, GLONASS. BDS 10/6/2019 07:56:59 (GMT -07:00) N 0.539444 w 0.244722 224
20191006_075705.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 10/6/2019 07:57:05 (GMT -07:00) N 0.539444 w 0.244722 224
20191006_075707.jpg Samsung Galaxy S6 A-GPS, GLONASS, BDS 10/6/2019 07:57:07 (GMT -07:00) N 0.539444 w 0.244722 224
IMG 2030-Airplane.JPG iPhone 7 Plus A-GPS, GLONASS. GALILEO, QZSS 10/6/2019 7:53:28 AM (Unknown) N w -
IMG 2031-Airplane.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:53:36 (GMT -07:00) N 0.539414 w 0.244575 215
IMG 2032.JPG iPhone 7 Plus A-GPS, GLONASS. GALILEO, QZSS 10/6/2019 07:54:01 (GMT -07:00) N 0.539322 w 0.244622 224
IMG_2033.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:54:12 (GMT -07:00) N 0.539331 w 0.244622 224
IMG_2034.JPG iPhone 7 Plus A-GPS, GLONASS, GALILEO, QZSS 10/6/2019 07:54:27 (GMT -07:00) N 0.539336 w 0.244636 224
IMG_0164.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:55:01 (GMT-07:00) N 32.359’ w 14.681’ 220.98
IMG_0165.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:55:02 (GMT -07:00) N 32.359’ w 14.681’ 220.98
IMG 0166.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:55:17 (GMT -07:00) N 32.359’ w 14.681’ 221.285
20191006 075705.jpg Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:57:05 (GMT -07:00) N 32.359’ w 14.680’ 217.018
20191006 075707.jpg Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:57:07 (GMT -07:00) N 32.359’ w 14.680’ 217.322
IMG 2033JPG Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:54:12 (GMT -07:00) N 32.358’ w 14.680’ 220.37
IMG_2034.JPG Garmin eTrex 20x Garmin Default + GLONASS 10/6/2019 07:54:27 (GMT -07:00) N 32.359’ w 14.680' 220.675
IMG_0162-Airplane.JPG GPSmap 62s Garmin Default 10/6/2019 07:54:54 (GMT -07:00) N 32*22.2" w 14’36.6" 222.199
IMG 0163-Airplane.JPG GPSmap 62s Garmin Default 10/6/2019 07:54:55 (GMT -07:00) N 32’22.2" w 14’36.6” 222.199
20191006 075656.jpg GPSmap 62s Garmin Default 10/6/2019 07:56:56 (GMT -07:00) N 3222.1" w 14’36.5" 220.675
20191006_075658.jpg GPSmap 62s Garmin Default 10/6/2019 07:56:57 (GMT -07:00) N 3222.1" w 14’36.5" 221.285
20191006 075659 jpg GPSmap 62s Garmin Default 10/6/2019 07:56:59 (GMT -07:00) N 32’22.2" w 14’36.6" 221.59
IMG_2032.JPG GPSmap 62s Garmin Default 10/6/2019 07:54:01 (GMT -07:00) N 32’22.4" w 14’36.4" 220.37
From Table 22, it is displayed that mostly all devices provided an elevation and GPS profile. It was discovered for this test that the iPhone Camera App option for location needed to have the location setting set to ‘Ask next time’. This prompts the app to ask the user if location is desirable for the users’ app session. This seems to enable GPS right away when enabled, as the device was mostly receiving GPS coordinates right after exiting that prompt.
The only anomaly with this location setting was from the iPhone 7 Plus ‘IMG 2030-Airplane.jpg’, which did not provide a GPS coordinate or elevation profile. However, a GPS coordinate was produced 8 seconds later with TMG_2031-Airplane.jpg’. With ‘IMG 2030-Airplane.jpg’ it was interesting to see that the time offset was not listed like the other photos EXIF data, as it produced an ‘unknown’ offset.
40


Within this location test the GPSmap 62s provided the worse average distance from actual location followed by the Samsung Galaxy S6, which provided the worse distance for a cellular device.
It was surprising that after the location setting switch to ‘Ask next time’ devices produced good results in the distance category. As the iPhone 6s received an average distance of 3.1061 meters from actual location and the iPhone 7 Plus received an average distance of 2.36365 meters from actual location. The Garmin eTrex 20x provided the best average distance of 1.5547 meters from actual location. Reference Figure 27 for all image GPS coordinates compared to the actual GPS survey marker location.
Figure 27: NGS Location II Urban #V -Image GPS Coordinates vs Actual Location
(SARTOPOMap)
41


Figure 28: NGS Location II Urban #V -Image GPS Coordinates vs Actual Location (Google
Earth Map)
With elevation the Samsung Galaxy S6 provided the best average elevation percent error of 0.44%. Surprisingly the Garmin eTrex 20x provided the worse elevation percent error of 2.31%. It was interesting to see that all devices did well in both elevation and distance compared to previous data sets. Figure 27 displays a Google Earth map to show the type of elevation associated to the environment type.
Table 22: NGS Location MI Urban #V - Percent Error Elevation and Average Distance from
Actual
File Name Device % Error Elevation Average % Error Elevation Distance From Actual (m) Average Distance from Actual (m)


IMG 0162-Airplane.JPG iPhone 6s 1.33% 7.634
IMG 0163-Airplane.JPG iPhone 6s 1.33% 7.634
IMG 0164.JPG iPhone 6s 1.78% 3.0639
IMG 0165.JPG iPhone 6s 1.78% 3.0639
IMG 0166.JPG iPhone 6s 0.89% 1.42% 3.1483 3.1061
20191006 075656.jpg Samsung Galaxy S6 0.44% 15.2978
20191006 075658.jpg Samsung Galaxy S6 0.44% 15.2978
20191006 075659.jpg Samsung Galaxy S6 0.44% 15.2978
20191006 075705.jpg Samsung Galaxy S6 0.44% 15.2978
20191006 075707.jpg Samsung Galaxy S6 0.44% 0.44% 15.2978 15.2978
IMG 2030-Airplane.JPG iPhone 7 Plus
IMG 2031-Airplane.JPG iPhone 7 Plus 4.44% 12.3639
IMG 2032.JPG iPhone 7 Plus -0.44% 11.8774
IMG 2033.JPG iPhone 7 Plus -0.44% 2.2505
IMG 2034.JPG iPhone 7 Plus -0.44% 0.78% 2.4768 2.36365
IMG 0164.JPG Garmin eTrex 20x 1.79% 2.5701
IMG 0165.JPG Garmin eTrex 20x 1.79% 2.5701
IMG 0166.JPG Garmin eTrex 20x 1.65% 2.5701
20191006 075705.jpg Garmin eTrex 20x 3.55% 1.2461
20191006 075707.jpg Garmin eTrex 20x 3.41% 1.2461
IMG 2033.JPG Garmin eTrex 20x 2.06% 2.1721
IMG 2034.JPG Garmin eTrex 20x 1.92% 2.31% 1.2461 1.55476667
IMG 0162-Airplane.JPG GPSmap 62s 1.24% 91.8726
IMG 0163-Airplane.JPG GPSmap 62s 1.24% 91.8726
20191006 075656.jpg GPSmap 62s 1.92% 93.3716
20191006 075658Jpg GPSmap 62s 1.65% 93.3716
20191006 075659.jpg GPSmap 62s 1.52% 91.8726
IMG 2032.JPG GPSmap 62s 2.06% 1.61% 97.4642 93.3042
42


IV. FINDINGS AND RESULTS EXPLANATION
Below will describe findings relating to distance and elevation error depending on device and environment type. The findings between Apple and Samsung device will the discussed. Observations with airplane and other notable findings will be addressed.
Distance Error
By acquiring the average distance in meters from each photo taken against the known actual location using the Matlab script: [arclen, az] = distance(lat1,Ionl,lcit2,lon2)*1000. Figure 29 displays each test location in respect to each device that was used. An overall average distance was calculated to display another figure of accuracy between all tests associated to the device. Additionally, an overall average distance for each environment type per each device was calculated and displayed to show which device excelled in different environment types. Below will explain what was observed in the device and environment types regarding distance.
iPhone 6s IPho te 7 iPhone 7 Plus iPhone 8
Location Avg Distance From Actual (m) Location Avg Distance From Actual (m) Location Avg Distance From Actual (m) Location Avg Distance â–  10-1 Ai a I -
Test #1 Rural #1 1825.30175 Test #1 Rural #1 Test #1 Rural #1 12.705175 Test#1 Rural #1
Test #2 Rural #2 24.112225 Test #2 Rural #2 Test #2 Rural #2 5446.5903 Test #2 Rural #2
Test #3 Rural #3 41.5166 Test #3 Rural #3 Test #3 Rural #3 32.6802 Test #3 Rural #3
Test #4 Rural #4 15.4937 Test #4 Rural #4 Test #4 Rural #4 17.91668333 Test #4 Rural #4
Test #5 Suburb #1 23.786675 Test #6 Suburb #1 Test #5 Suburb #1 9.245175 Test #5 Suburb #1
Test #6 Suburb #2 Test #6 Suburb #2 Test #6 Suburb #2 8.4002 Test #6 Suburb #2
Test #7 Suburb #3 | 8.408214286 Test #7 Suburb #3 Test #7 Suburb #3 13.57837143 Test #7 Suburb #3
Test #8 Suburb #4 27.32176667 Test #8 Suburb #4 Test #8 Suburb #4 18.452025 Test #8 Suburb #4
Test #9 Urban #1 15.6929 Test #9 Urban #1 74.0605 Test #9 Urban #1 111.03935 Test #9 Urban #1 11.8057
Test #10 Urban #2 2.7142 Test #10 Urban #2 Test #10 Urban #2 134.33735 Test #10 Urban =2
Test #11 Rural #5 3.1061 Test #11 Rural #5 Test #11 Rural #5 2.36365 Test #11 Rural #5

Overall 198.7454131 Overall 74.0605 Overall 527.9371345 Overall 11.8057
Overall Rural 381.906075 Overall Rural Overall Rural 1102.451202 Overall Rural
Overall Sub 19.83888532 Overall Sub Overall Sub 12.41894286 Overall Sub
Overall Urban 9.20355 Overall Urban 74.0605 Overall Urban 122.68835 Overall Urban 11.8057
Samsung Galaxy S6 Samsung SM-N960U Garmin GPSmap 62s Garmin Etrex 20x
Location Avg Distance From Actual (m) Location Avg Distance From Actual (m) Location Avg Distance From Actual fm) Location Avg Distance From Actual (m)
Test #1 Rural #1 Test #1 Rural #1 12.5108 Test #1 Rural #1 79.5712 Test #1 Rural #1 13-2867
Test #2 Rural #2 Test #2 Rural #2 Test #2 Rural #2 86.509675 Test #2 Rural #2 29.761025
Test #3 Rural #3 0 Test #3 Rural #3 Test #3 Rural #3 88.249075 Test #3 Rural #3 13.472775
Test #4 Rural #4 0 Test #4 Rural #4 Test #4 Rural #4 79.4078 Test #4 Rural #4 10.6097
Test #5 Suburb #1 Test #5 Suburb #1 3804.9461 Test #5 Suburb #1 96.99945 Test #5 Suburb #1 6.8613333
Test #6 Suburb #2 Test #6 Suburb #2 21.404875 Test #6 Suburb #2 85.9497667 Test #6 Suburb #2 3.6817
Test #7 Suburb #3 0 Test #7 Suburb #3 Test #7 Suburb #3 100.5164667 Test #7 Suburb #3 6.55045
Test #8 Suburb #4 38.0141 Test #8 Suburb #4 Test #8 Suburb #4 124.554 Test #8 Suburb #4 6.437575
Test #9 Urban #1 0 Test #9 Urban #1 Test #9 Urban #1 109.42566 Test #9 Urban #1 11.68553333
Test #10 Urban #2 29.8002 Test #10 Urban #2 Test #10 Urban #2 117.7073 Test #10 Urban #2 26.8196333
Test #11 Rural #5 15.2978 Test #11 Rural #5 ! r.-i =â–  93.3042 Test #11 Rural #5 1.55476667

Overall 11.87315714 Overall 1279.620592 Overall 96.56314485 Overall 11.88374469
Overall Rural 5.099266667 Overall Rural 12.5108 Overall Rural 85.40839 Overall Rural 13.73699333
Overall Sub 19.00705 Overall Sub 1913.175488 Overall Sub 102.0049209 Overall Sub 5.882764575
Overall Urban 14.9001 Overall Urban Overall Urban 113.56648 Overall Urban 19.25258332
Figure 29: Distance Error from Actual Location
43


Device
From data in Figure 29, it can be shown that the Samsung Galaxy S6, iPhone 8 and the Garmin Etrex 20x displayed an average distance closest to the actual location (0). However, the iPhone 8 was only used in one experiment, so it cannot be proven that it displayed a better GPS accuracy. The next runner up would be the Samsung Galaxy S6, which makes sense merely that on some tests the phone displayed no distance error.
The worse devices displayed from Figure 29 were the Samsung SM-N960U and the iPhone 7 Plus. These were most likely due to anomalies of GPS locations putting the devices in very far out places when trying to secure a promising GPS coordinate. Two instances, one from each device showed this anomaly and will be explained in the other findings section later throughout the paper. However, I believe these two outliers contributed to these devices performing poorly by providing the overall worse average distance from actual location.
Environment
Apple
Avg Distance From Actual (m)
Overall 203.1371869
Overall Aural 742.1786383
Overall Sub 16.12891409
Overall Urban 43.55162
Garmin GPS Units
Avg Distance From Actual (m)
Overall 54.22344477
Overall Rural 49,57268167
Overall Sub 53.94384271
Overall Urban 113.56648
Samsung
Avg Distance From Actual (m)
Overall 645.7468744
Overall Rural 12.5108
Overall Sub 966.0912688
Overall Urban 14,8001
Environment All Devices)
Rural 225.7768765
Sub 201.06SS32S
urban 57,3394
Apple + Samsung
Distance From Actual (mi
Overall 424.4420307
Overall Rural 377,3447192
Overall Sub 491.1100914
Overall Urban 29.22586
Figure 30: Device Average Distance Pertaining to Environment
From Figure 30, Garmin devices appear to have the most accurate distances. Apple being second closest and Samsung being the farthest away from the actual. However, these differ from
44


each type of environment. This figure also paints a picture of how these environment types vary as far as accuracy all together.
Looking the overall rural figure for both companies. Apple’s overall rural figure is 742.1786 meters from actual location and Samsung’s overall rural figure is 12.5108 meters from actual location. But in a suburb environment these two companies both switched roles. Samsung’s overall suburb figure being 966.091 meters from actual location to Apple’s overall suburb figure to 16.1289 meters from actual location. Below will outline each environment type for the best and worse devices for each setting.
Rural
From Figure 30 and looking at overall rural results. Samsung devices appear to be closest to the actual location, with Garmin GPS units following and Apple devices being the farthest from actual location. Apple and Samsung devices collectively have a rural overall average distance of 377.3447192 meters. All devices in this environment have an average distance of 225.7768765 meters.
Figure 31 displays a scatter plot of all rural location tests compared to the actual location point. A trend of plots staying consistent with a longitude point ‘-122.2446’ is shown. It seems the latitude point in this graph is the major defining factor of where that point would lie in relation to the actual point plot.
45


Rural Location Plots vs Actual
-122.2432
45.53915 45.5392 45.53925 45.5393 45.53935 45.5394 45.53945 45.5395 45.53955 45.5396 45.53965
-122.2434 -------------------------------------------------•----------------------•-------•-------
• • •
-122.2436 ----------------------------------------------------------------------------------------
-122.2438 ----------------------------------------------------------------------------------------
(LI
T3
13 Actual Location Plot
-122.2448 --------------------------------------------------
Latitude

Figure 31: Rural Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph
Suburb
From Figure 30 and looking at overall suburb results. Apple devices appear to be closest to the actual location, with Garmin GPS units following and Samsung devices being the farthest from actual location. Apple and Samsung devices collectively have a suburb overall average distance of 491.1100914 meters. All devices in this environment have an average distance of 201.0655325 meters.
Figure 32 displays a scatter plot of all suburb location tests compared to the actual location point. Again a trend is displayed of the longitude staying consistent for the actual point plot longitude. Latitude has a gap displayed from approximately ‘45.5303’ - ‘45.5310’.
46


Suburb Location Plots vs Actual
-122.437 |-------------------------1------------------------------------------------------------------------
45.3302 45.5303 45.5304 45.5305 45.5306 45.5307 45.530S 45.5309 45.531
-122.4375 |
-122.438
-122.4385
07
â– o
~ -122.439 o
_i
-122.4395
-122.44
-122.4405
-122.441
Actual Location Plot
• •
• •
** H* JM
Latitude
Figure 32: Suburb Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph
Urban
From Figure 30 and looking at overall urban results. Samsung devices appear to be closest to the actual location, with Apple devices following and Garmin GPS units being the farthest from actual location. Apple and Samsung devices collectively have an urban overall average distance of 29.22586 meters. All devices in this environment have an average distance of 57.3394 meters.
Figure 33 displays a scatter plot of the urban location test compared to the actual location point. This provides a nice visual of the inconsistencies that arose from the different devices and what trends were demonstrated. From previous scatter plots graph the same trend that devices seem to follow are shown.
47


Urban Location Plots vs Actual
-122.6784
-122.67S&5' 524
-122.6788 -122.679 -122.6792

3 -122.6794
“ -122.6796 o
-122.6798
-122.68
-122.6802
-122.6804
-122.6806
Figure 33: Urban Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph
Elevation Error
Elevation Error was determined by using the percent error formula displayed in Figure 4. This was used to come up with a percent error for each elevation data point from each test image. An average percent error from each device test was plotted in Figure 34. An overall percent error for each environment was calculated and displayed in Figure 34.
iPhone 7 Plus iPhone 8
Location Avg % Error Location Avg % Error
Test #1 Rural #1 1.80% Test #1 Rural #1
Test #2 Rural #2 27.74% Test #2 Rural #2
Test #3 Rural #3 0.28% Test #3 Rural #3
Test #4 Rural #4 44.66% Test #4 Rural #4
Test #5 Suburb #1 1.57% Test #5 Suburb #1
Test #6 Suburb #2 1.57% Test #6 Suburb #2
Test #7 Suburb #3 3.23% Test #7 Suburb #3
Test #8 Suburb #4 5.27% Test #8 Suburb #4
Test #9 Urban #1 64.85% Test #9 Urban #1 3.81%
Test #10 Urban #2 330.69% Test #10 Urban #2
Test #11 Rural #5 0.78% Test #11 Rural #5
iPhone 6s
Location Avg % Error
Test #1 Rural #1 16.34%
Test #2 Rural #2 1.21%
Test #3 Rural #3 8.78%
Test #4 Rural #4 3.50%
Test #5 Suburb #1 4.13%
Test #6 Suburb #2 -
Test #7 Suburb #3 7.26%
Test #8 Suburb #4 4.86%
Test #9 Urban #1 10.63%
Test #10 Urban #2 14.40%
Test #11 Rural #5 1.42%
Overall 7.25%
Overall Rural 6.25%
Overall Sub 5.42%
Overall Urban 12.52%
Overall 3.81%
Overall Rural
Overall Sub
Overall Urban 3.81%
Overall 43.86%
Overall Rural 15.05%
Overall Sub 2.91%
Overall Urban 197.77%
48


Garmin Etrex 20x
Location Avg % Error
Test #1 Rural #1 6.21%
Test #2 Rural #2 0.84%
Test #3 Rural #3 2.32%
Test #4 Rural #4 2.23%
Test #5 Suburb #1 1.00%
Test #6 Suburb #2 22.50%
Test #7 Suburb #3 43.03%
Test #8 Suburb #4 9.60%
Test #9 Urban #1 75.97%
Test #10 Urban #2 24.10%
Test #11 Rural #5 2.31%
Samsung SM-N960U
Location Avg % Error
Test #1 Rural #1 6.25%
Test #2 Rural #2 -
Test #3 Rural #3 -
Test #4 Rural #4 -
Test #5 Suburb #1 -
Test #6 Suburb #2 40.94%
Test #7 Suburb #3 -
Test #8 Suburb #4 -
Test #9 Urban #1
Test #10 Urban #2 -
Test #11 Rural #5 -
Samsung Galaxy S6
Location Avg % Error
Test #1 Rural #1
Test #2 Rural #2
Test #3 Rural #3 3.45%
Test #4 Rural #4 15.35%
Test #5 Suburb #1 90.60%
Test #6 Suburb #2
Test #7 Suburb #3 38%
Test #8 Suburb #4 34%
Test #9 Urban #1 100%
Test #10 Urban #2 100%
Test #11 Rural #5 0.44%
Garmin GPSmap 62s
Location Avg % Error
Test #1 Rural #1 5.93%
Test #2 Rural #2 8.72%
Test #3 Rural #3 2.52%
Test #4 Rural #4 1.69%
Test #5 Suburb #1 3.83%
Test #6 Suburb #2 23.50%
Test #7 Suburb #3 65.50%
Test #8 Suburb #4 2.50%
Test #9 Urban #1 363.72%
Test #10 Urban #2 31.63%
Test #11 Rural #5 1.61%
Overall 17.28%
Overall Rural 2.78%
Overall Sub 19.03%
Overall Urban 50.04%
Overall 23.60%
Overall Rural 6.25%
Overall Sub 40.94%
Overall Urban
Overall 46.47%
Overall Rural 4.09%
Overall Sub 23.83%
Overall Urban 197.68%
Overall 47.74%
Overall Rural 6.41%
Overall Sub 54.21%
Overall Urban 100.00%
Figure 34: Overall Elevation Percent Error
Device
Figure 35 ranks each device from overall lowest average elevation percent error.
Elevation Average Percent Error
Device Avg% Error
iPhone 6s 7.25%
Garmin Etrex 20x 17.28%
Samsung SM-N960U 23.60%
iPhone 7 Plus 43.86%
Garmin GPSmap 62s 46.47%
Samsung Galaxy S6 47.74%
iPhone 7 _^-*Kr7T%
Figure 35 : Elevation Average Percent Error Device Ranking
Both the iPhone 8 and iPhone 7 have a red line through their results, due to both devices only being tested in one experiment. These devices cannot be shown as either the best or worse device in respect to average percent error for all tests and will not be taken account for the ranking.
Glancing over each device shows that there is no constant percent error rate relating to each environment. There is much of a change depending on the device itself and not solely on any environment type.
49


Environment
From Figure 34, each environment had different devices that excelled over others. Within this section each environments percent error for devices will be addressed.
In a rural setting the Garmin eTrex had the lowest percent error. For cellular devices, the iPhone 6s had the lowest percent error. The worse device in this category was the iPhone 7 Plus. Having a Garmin unit providing the lowest percent error rate in this test was not too surprising as this is a stand-alone GPS unit and its main function is to be able to provide ultimate accuracy in this type of environment setting.
In a suburb setting the iPhone 7 Plus had the lowest percent error. The worse device in this category was the Samsung Galaxy S6. It was interesting that the Apple devices did not produce similar results overall. Since they are pretty on par in respects to the device models not having a huge generation gap difference.
In an urban setting the iPhone 8 had the lowest percent error but was only used in one test. This leads to the runner up, the iPhone 6s having the lowest percent error. The worse device in this category was the Garmin GPSmap 62s. It was intriguing to see that in a urban environment the Garmin devices did not excel, which might be due to other interference of devices and buildings that encompass an urban setting.
Apple vs Samsung
Apple and Samsung have varied results when compared to overall average distance from an actual location and different environments. Figure 36 displays each environment with an overall distance from an actual location. Apple devices are favored overall in accuracy from the
50


tests in acquiring an average distance from an actual location and overall in a suburb environment. Where Samsung devices are favored in a rural and urban environment.
Apple
Avg Distance From Actual (m)
Overall . . â– ;
Overall Rural 742.1786383
Overall Sub 16.12891409
Overall Urban 43.55162
Samsung
Avg Distance From Actual (m)
Overall 645.7468744
Overall Rural 12.5108
Overall Sub 966.0912688
Overall Urban 14.9001
Figure 36: Apple us Samsung Distance from Actual
This is an interesting observation since these two companies use 3 similar satellites, however different in one. Maybe the BDS satellite system that Samsung utilizes favors rural and urban environments over suburbs? Or the same could be said over Apple’s QZSS satellite systems favoring suburb environments over rural and urban. It would make sense that these results would be different because one satellite system, in theory, should mean different results between these two companies.
Airplane Mode
From Test series #1, each cellular device followed the procedure of cell service on, then switching to airplane mode, then a photo being captured. However, it was then realized that the cellular device might be keeping a known location within the cache of the phone. This suggests after having the device change to airplane mode, the device may rely on the previous known location. Which in theory, the previous known location would be attached to the metadata of any new images.
Test series #11 were tests focused on how airplane mode alters GPS metadata to photos and if tests can prove that a cellular device might retain known locations from previous photos.
51


From referencing section, Tests series #11: NGS Locations and Initial Airplane Mode Tests, the phones were first put into airplane mode, powered off and then turned back on ensuring airplane mode was still active. From tests administered, this revealed the phone does still capture GPS metadata, depending on device location settings.
For Apple Devices, a user has three options to choose from under location services for capturing locations from the device’s camera app. These three options are to allow location access to the camera app by either ‘never’, ‘ask next time’ or ‘while using the app’. Under NGS Location #1, the iPhone 6s and iPhone 7 Plus did not initially provide a photo GPS coordinate, but the next photo did provide a GPS coordinate. This is because the ‘while using the app’ option takes roughly 30 secs for the Camera app to trigger the GPS in the device to acquire a location for the photo. However, in NGS Location #11, when testing the ‘ask next time’ option. This option triggers the camera app to regularly turn on GPS right away displaying GPS coordinates in photo metadata.
For Samsung devices, three options are available to choose under location settings. These three options are ‘high accuracy’, ‘battery saving’ and ‘device only’. Under Test series #11, the ‘device only’ option was utilized and made the test solely based on the devices GPS. This process did take about five minutes for the device to acquire a GPS coordinate. However, a GPS coordinate was produced within photo metadata.
Other Notable Findings
Random anomalies would occur with some of the GPS metadata associated to the photos. Within this section a discussion of notable anomalies will be addressed.
52


One in particular was in Test #1 Rural Location #11, with the iPhone 7 Plus. The test environment was rural and the first photo on that device displayed a coordinate that was 21717.1 meters away from the actual location. This could have been from the device taking some time to catch up to acquire a GPS satellite. However, most of the time this type of encounter that was this far of a distance in tests administered were not seen.
Another anomaly had to do with Test #1 Suburb Location #1, with a Samsung SM-N960U device. The interest in this anomaly shows a total of four photos taken within this test series.
The first three photos were near the actual location of the test series. However, the last photo brought the GPS coordinate all the way to a previous test location that was 15,164.6 meters away. Speculation thinks that maybe the phone lost connection to a previous GPS satellite and only fell back to a previous known coordinate. Further cellular device forensics would have to be conducted to determine any other background processes causing this change.
Previously noted before in this paper, the Samsung Galaxy S6 with no cell service activated will take roughly 45 secs to 5 mins to acquire a GPS coordinate. However, that coordinate most of the time was accurate to the actual location being tested. My assumption is that other connection types i.e. cell service, WIFI and Bluetooth could in fact hinder the device to determine a precise location. It was assumed that a device also utilizing these other connections would provide accuracy to the device. This could differ for each device model or software update and will need further testing.
During tests, a situation came up when a photo taken in a rural location was displaying GPS coordinates of a residential area. The Samsung device had the location setting of ‘high accuracy’ implemented, which means that it will acquire a location from all connection types available. After some research of the GPS coordinate, it was determined that the device pulled GPS
53


coordinates from the connection to a WIFI router. Further cellular forensics would have to be
done to explain any background processes occurring. But it is strange for this the cellular device to provide a location in this fashion.
54


V.
FUTURE RESEARCH
Much more future research could be conducted on this topic focusing many different types of details photo EXIF data exhibits.
Analysis of each device and operating system software versions from each cellular company could be a study to see if one operating system differs from another. It is possible for device architecture or software updates to be different and could change the priority list of how a device is determining a location. Also, a software update could maybe implement a new technology for the device to utilize connections more efficiently.
Types of weather tests could further be done to determine if this causes GPS interferences. Maybe there can be a trend identified to determine any error offsets that could be done to account for this type of circumstance.
Tests being catered around the anomalies that were being experienced throughout the paper would be another good research study. Mainly around what specific parameters are taking place in different devices and to explain how a device will output a certain GPS coordinate. It would probably be clear to have a forensic download of each cellular device and analyze what processes the phone is going through when an anomaly occurs.
Being able to pinpoint what GPS satellites or satellite systems a device is talking to would be interesting to see if a device is for certain connecting to the closest satellites. This eludes from previous statements focusing on further research between Samsung’s BDS satellite system difference over Apple’s QZSS satellite system.
Diving deeper into tests involving airplane mode would be another option. It was touched briefly with the last two tests administered and could be expanded to acquire more of overall
55


determination of what processes may be happening in the background. Test on solely WIFI GPS coordinates would be beneficial to see the accuracy of device is a factor to known WIFI router locations. With the implementation of 5Gthis could become a trend in GPS accuracy.
56


VI. CONCLUSION
This study came forth with the intention to determine how accurate photo GPS data is from cellular devices compared to GPS coordinates from a standalone GPS unit in different environments. It was administered in two different ways to determine location accuracy, utilize calibrated known location coordinates and test airplane mode possible restrictions to provide a valid data set.
Distance errors, elevation errors and anomalies were addressed. When these errors were compared to devices to see what trends or accuracy could be revealed. It was mainly addressed that device accuracies depended on different factors. The type of environment, cell service and other functions that may be going on in the background of the device could contribute to the cause of these anomalies. It cannot be said if one device is more accurate than the other for this reason.
Satellite systems that Apple and Samsung use within their device architecture were addressed and provided results that displayed different GPS coordinates and elevation profiles. Suggesting that Apple and Samsung satellite choice may excel in different environment types and each device could be useful in different ways depending on the environment.
Airplane mode was experimented in administering tests. The first series of tests had devices with cell service active first then switched over to airplane mode. The second series of tests had the device powered off with airplane mode and powered back on with airplane mode active. It was found that depending on the device location settings, determines how fast the device may acquire a GPS coordinate. Both series of tests types and still produced results that led to a GPS coordinate being created. However, there is still a chance that after a device has been powered on
57


and initial photos taken. There will not be a GPS coordinate produced depending on location settings used.
Environment tests were administered and does play a role in devices being able to acquire an accurate GPS coordinate and elevation profile. Throughout tests it was displayed that devices favored an urban environment for GPS accuracy and least favored a rural environment. Many anomalies were involved in each test and seemed each device exceled in certain environments differently than others.
Ultimately, many considerations come into play with GPS accuracy on smartphones. From research, it should be considered that claims made from GPS data in image files should be examined carefully. All factors of cell service, WIFI connection, environment and device models play a role of an overall GPS coordinate created from a device. A GPS coordinate could be accurate or can have a random anomaly applied to it due to one of these factors. It has been displayed that different outputs of GPS coordinates are produced by different devices. Additionally, these outputs can be produced by simply having a cellular device in airplane mode. Different devices excel in different types of environments relating to rural, suburb and urban areas.
Photo GPS data can be very beneficial in an investigation, especially if other forensic practices or key information of an investigation support any findings from GPS image data.
From different anomalies previously addressed, having a validation standard implemented would be beneficial for the digital forensics community to have best practices set into place for use of EXIF image data. Even having a clearinghouse of cellular devices being tested against current GPS satellites would be a benefit for the digital forensic community. This type of GPS data
58


within images will become even more popular with new technologies being implemented in our way of life.
59


REFERENCES
Exiv 2. Metadata Reference Tables, n.d. 15 Oct 2019. .
Explanation, Garmin-GPS. https://www8.garmin.com/aboutGPS/. 2017. 15 September 2019.
Garmin. Garmin ETREX® 10/20/2Ox/30/3Ox Owners Manual. Kansas: Garmin ltd, 2019.
Garmin. Garmin Manual GPSmap 62s. Kansas: Garmin ltd, 2010-2011.
gps.gov. https://www.gps.gov/systems/gnss/. 18 December 2017. 15 September 2019.
Merry, Krista, and Pete Bettinger. "SmartPhone GPS accuracy study in an urban environment." PloSone Vol 14,7 (2019). .
National Geodetic Survey. Survey Maks and Datasheets. n.d. 16 October 2019. .
Sandoval Orozco, A.L., Arenas Gonzalez, D.M., Garcia Villalba. "Analysis of errors in exif metadata on mobile devices." Multimed Tools Appl (2015). .
Zamir, Amir Roshan. "Visual Geo-Localization and Location-Aware Image Understanding." University of Central Florida (2014). .
60


Full Text

PAGE 1

ACCURACY OF IMAGE GPS EXIF DATA FROM APPLE AND SAMSUN G MOBILE DEVICES COMPARED TO GPS UNIT by DANIEL FELIX B.S., San Diego State University, 2013 A thesis submitted to the Faculty of Graduate School of the University of Colorado in partial fulfillment o f the requirements for the degree of Master of Science Recording Arts Program 2019

PAGE 2

ii This thesis for the Master of Science degree by Daniel Felix has been approved for the Recording Arts Program by Catalin Grigoras, Chair Jeff M. Smith Jason Lewis Cole Whitecotton Date: December 14, 2019

PAGE 3

iii Felix, Daniel (M.S., Recording Arts Program) Accuracy of Image GPS EXIF D ata from Apple and Samsung Mobile Devices Compared to GPS U nit Thesis directed by Associate Professor Catalin Grigoras ABSTRACT Mostly every photo captured from a cellular device contains EXIF GPS coordinates associated to the location of where that photo was captured. This thesis will p ropose tests for accuracy of GPS EXIF data from different Apple, Samsung and Garmin GPS devices . Apple has used the same GPS satellites in their cellular devices within the past 4 generations of models released . Samsung has used the same GPS satellites from the past 3 generations of models released . Both Apple and Samsung use a total of 4 GPS satellites, but differ in one. The proposed tests will determine if this one different satellite may cause separate results to be produced. Within this expe riment two types of tests will be administered. The first test will involve a focus of a cellular device capturing photos by having cell service on , then switching to airplane mode. The second test will involve a fo cus of a cellular device capturing photos by having airplane mode on initially then switching to cell service active . Results of both tests will be analyzed and any anomalies will be addressed. NGS survey markers will be utilized to explore the idea of a more prominent point established when gathering test information. Image GPS data from cellular devices will be compared to a standalone GPS device readings . Experiments will take place in urban and rural environments and results will be analyzed. The f orm and content of this abstract are approved. I recommend its publication . Approved: Catalin Grigoras

PAGE 4

iv This thesis is dedicated to my family. For always supporting me in any aspirations I wanted to pursue and for guiding me t hroughout life.

PAGE 5

v ACKNOWLEDGMENTS A special thank you for my thesis committee, Catalin, Jeff, Jason and Cole for all the work and gu idance towards this thesis. Their support throughout this process was extremely appreciative and brought a great skillset to support this thesis. Catalin and Jeff, thank you for the past two years throughout the Media Forensic s graduate program at CU Denve r. You both have showed me skills that I will have the rest of my career and opened my eyes to a new passion of media forensics. Leah and Cole, thanks for the added support throughout the past two years. You both were easy to approach with any questions th at may arise and are an invaluable asset to the Media Forensic s program. To the Media Forensic s cohort 2017 19, thanks for all the support throughout our program and I am super grateful do this program with such a great group of individuals.

PAGE 6

vi TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ ............ 1 Literature Review ................................ ................................ ................................ .................... 2 II. METHODOLOGY ................................ ................................ ................................ ........... 4 Description of Materials (Devices Used) ................................ ................................ .................. 4 Software ................................ ................................ ................................ ................................ .. 4 Pic2Map ................................ ................................ ................................ .............................. 5 Google Earth ................................ ................................ ................................ ........................ 5 SARTOPO ................................ ................................ ................................ ........................... 5 JPEGSnoop ................................ ................................ ................................ .......................... 5 Matlab ................................ ................................ ................................ ................................ . 6 GPS Explanation ................................ ................................ ................................ ..................... 6 A GPS ................................ ................................ ................................ ................................ . 7 GLONASS ................................ ................................ ................................ ........................... 7 GALILEO ................................ ................................ ................................ ............................ 7 QZSS ................................ ................................ ................................ ................................ ... 7 BDS ................................ ................................ ................................ ................................ ..... 7 EXIF Photo GPS Explanation from JPEGSnoop ................................ ................................ ...... 8 Method Used to Acquire Data ................................ ................................ ................................ .. 9

PAGE 7

vii Test #I Locations #I IX ................................ ................................ ................................ ... 10 Test #II Locations #X #XI: NGS Locations and Initial Airplane Mode Tests ................... 11 Analysis to Determine El evation Error ................................ ................................ ............... 12 Analysis to Determine Distance from Actual Location ................................ ....................... 13 III. TESTING LOCATIONS AND RESULTS FROM TEST #I & #II ................................ .. 14 Test #I Rural Location #I ................................ ................................ ................................ .... 14 Test #I Rural Location #II ................................ ................................ ................................ ... 17 Test #I Rural Location #III ................................ ................................ ................................ . 20 Test #I Rural Location #IV ................................ ................................ ................................ . 23 Test #I Suburb Location #I ................................ ................................ ................................ . 25 Test #I Suburb Location #II ................................ ................................ ................................ 27 Test #I Suburb Location #III ................................ ................................ ............................... 29 Test #I Suburb Location #IV ................................ ................................ ............................... 32 Test #I Urban Location #I ................................ ................................ ................................ ... 34 Test #II NGS Location #I Urban #II ................................ ................................ .................... 36 Test #II NGS Location #II Rural #V ................................ ................................ ................... 38 IV. FINDINGS AND RESULTS EXPLANATION ................................ .............................. 43 Distance Error ................................ ................................ ................................ ....................... 43 Device ................................ ................................ ................................ ............................... 44 Environment ................................ ................................ ................................ ...................... 44

PAGE 8

viii Rural ................................ ................................ ................................ .............................. 45 Urban ................................ ................................ ................................ ............................. 47 Elevation Error ................................ ................................ ................................ ...................... 48 Device ................................ ................................ ................................ ............................... 49 Environment ................................ ................................ ................................ ...................... 50 Apple vs Samsung ................................ ................................ ................................ ................. 50 Airplane Mode ................................ ................................ ................................ ....................... 51 Other Notable Findings ................................ ................................ ................................ .......... 52 V. FUTURE RESEARCH ................................ ................................ ................................ ... 55 VI. CONCLUSION ................................ ................................ ................................ .............. 57 REFERENCES ................................ ................................ ................................ ......................... 60

PAGE 9

ix LIST OF TABLES TABLE Table 1: Test #I Rural Location #I Coordinates and Elevation ................................ ................ 15 Table 2: Rural Location #I Percent Error Elevations and Average Distance from Actual ........ 17 Table 3: Rural Location #II Coordinates and Elevation ................................ ........................... 18 Table 4: Rural Locat ion #II Percent Error Elevations and Average Distance from Actual ....... 20 Table 5: Rural Location #III Coordinates and Elevation ................................ ......................... 21 Table 6: Rural Location #III Percent Error Elevation and Average Distance from Actual ....... 22 Table 7: Rural Location #IV Rural #IV Coordinates and Elevation ................................ ......... 23 Table 8: Rural Location #IV Percent Error Elevation and Average Distan ce from Actual ...... 25 Table 9: Suburb Location #I Coordinates and Elevation ................................ ......................... 25 Table 10: Suburb Location #I Percent Error and Average Distance from Actual ..................... 27 Table 11: Suburb Location #II Coordinates and Elevation ................................ ...................... 28 Table 12: Test # VI Suburb #II Percent Error Elevation and Average Distance from Actual .... 29 Table 13: Suburb L ocation #III Coordinates and Elevation ................................ ..................... 30 Table 14: Suburb Location #III Percent Error Elevation and Distance from Actual ................. 31 Table 15: Suburb Location #IV Coordinates and Elevation ................................ ..................... 32 Table 16: Suburb Location #IV Percent Error and Average Distance from Actual ................. 33 Table 17: Urban Location #I Coordinates and Elevation ................................ ......................... 34 Table 18: Urban Location #I Percent Error Elevation and Average Distance from Actual ....... 36 Table 19: NGS Location #I Urban #II Coordinates and Elevation ................................ ........... 37 Table 20: NGS Location #I Urban #II Percent Error and Average Distance from Actual ....... 38 Table 21: NGS Location #II Rural #IV Coordinates and Elevation ................................ ......... 40

PAGE 10

x Table 22: NGS Location #II Urban #V Percent Error Elevation and Average Distance from Actual ................................ ................................ ................................ ................................ ....... 42

PAGE 11

xi LIST OF FIGURES FIGURE Figure 1: Devices Used ................................ ................................ ................................ ............... 4 Figure 2: EXIF GPS Example: JPEGSnoop ................................ ................................ ................. 9 Figure 3: EXIF Hex View (HxD) ................................ ................................ ................................ 9 Figure 4: Photo Example from Cell Phone Device ................................ ................................ ..... 11 Figure 5: Percent Error Formula ................................ ................................ ................................ 13 Figure 6: Rural Location #I Location (SARTOPO Map) ................................ ................................ ................................ ....... 15 Figure 7: Rural Location #I Image GPS Coordinates vs Actual Location (SARTOPO map) ... 16 Figure 8: Rural Location #I Image GPS Coordinates vs Actual Location (Google Earth map) 16 Figure 9: Rural Location #II Image GPS Coordinates vs Actual Location (SARTOPO map) .. 18 Figure 10: Rural Location #II Anomaly IMG_1950.jpg Distance from Actual Location (SARTOPO map) ................................ ................................ ................................ ...................... 19 Figure 11: Rural Location #II Anomaly #2 iPhone 7 Plus Distance Corrections to Actual Location (SARTOPO Map) ................................ ................................ ................................ ....... 19 Figure 12: Rural Location #III Image GPS Coordinates vs Actual Location (SARTOPO Map) ................................ ................................ ................................ ................................ ................. 22 Figure 13: Rural Location #IV Image GPS Coordinates vs Actual Location (SARTOPO Map) ................................ ................................ ................................ ................................ ................. 23 Figure 14: Rural Location #IV Image GPS Coordinates vs Actual Location (Google Earth Map) ................................ ................................ ................................ ................................ ......... 24

PAGE 12

xii Figure 15: Suburb Location #I Distance from Ac tual Location (SARTOPO Map) ................................ ................................ ..... 26 Figure 16: Suburb Location #I Image GPS Coordinates vs Actual Location (SARTOPO Map) ................................ ................................ ................................ ................................ ................. 26 Figure 17: Suburb Location #I Image GPS Coordinates vs Actual Location (Google Earth Map) ................................ ................................ ................................ ................................ ......... 27 Figure 18: Suburb Location #II Image GPS Coordinates vs Actual Location (SARTOPO Map) ................................ ................................ ................................ ................................ ................. 28 Figure 19: Suburb Location #II Image GPS Coordinates vs Actual Location (Google Earth Map) ................................ ................................ ................................ ................................ ......... 29 Figure 20: Suburb Location #III Image GPS Coordinates vs Actual Location (SARTOPO Map) ................................ ................................ ................................ ................................ ................. 31 Figure 21: Suburb Location #IV GPS Coordinates vs Actual Location (SARTOPO Map) ...... 33 Figure 22: Urban Location #I Image GPS Coordinates vs Actual Location (SARTOPO Map) 35 Figure 23: Urban Location #I Image GPS Coordinates vs Actual Location (Google Earth Map) ................................ ................................ ................................ ................................ ................. 35 Figure 2 4: NGS Location #I Urban #II NGS Data Sheet Rural Environment and Survey marker 54 RESET ................................ ................................ ................................ ................................ . 37 Figure 25: NGS Location #I Urban #II Image GPS Coordinates from Actual Location (SARTOPO Map) ................................ ................................ ................................ ..................... 38 Figure 26: NGS Location #II Urban #V NGS Data Sheet Urban Environment and Survey Marker RD 2197 ................................ ................................ ................................ ....................... 39

PAGE 13

xiii Figure 27: NGS Location #II Urban #V Image GPS Coordinates vs Actual Location (SARTOPO Map ) ................................ ................................ ................................ ..................... 41 Figure 28: NGS Location #II Urban #V Image GPS Coordinates vs Actual Location (Google Earth Map) ................................ ................................ ................................ ................................ 42 Figure 29: Distance Error from Actual Location ................................ ................................ ........ 43 Figure 30: Device Average Distance Pertaining to Environment ................................ ................ 44 Figure 31: Rural Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph .............. 46 Figure 32: Suburb Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph ............ 47 Figure 33: Urban Distance Image GPS Plot vs Actual Location Plot Scatterplot Graph ............. 48 Figure 34: Overall Elevation Percent Error ................................ ................................ ................ 49 Figure 35 : Elevation Average Percent Error Device Ranking ................................ ................... 49 Figure 36: Apple vs Samsung Distance from Actual ................................ ................................ .. 51

PAGE 14

1 I. INTRODUCTION Photo EXIF data has become popular in the past couple of yea rs to the general public . The illusion of a photo being simply a photo is not the case anymore . T he data that can be captured by these image files can encompass personal information of a user . Within a photo, GPS metadata can be captured that contains the exact coordinates of where a person was previously located . This brings up question s of how a cellular device is creating an image file with this background information and the accuracy of this data . Within this paper , the accuracy of EXIF data being capture d from Apple and Samsung cellular devices will be compared to GPS standalone devices. Airplane mode is another feature within cellular devices that will be explored . This feature turns off cell service associated to the device . T his function will be experimented on to see if GPS data is still captured a nd outputted to image metadata . Tests will be conducted in three different environments to determine if accuracy is swayed. Two tests will be done using NGS ( National Geodetic Survey ) marker locations to see if any different outcomes were noticed. Tests w ill be analyzed and compared to determine any interesting trends and findings. These trends could be associated to environment, airplane mode, elevation and phone manufacturer associated to the device. Any anomalies will be addressed and analyzed to determ ine the reason an anomaly occurred. This thesis will be organized by chapter s and will outline the steps of the overall test process. Chapter I presents the introduction and literature review. Chapter II presents the methodology describing software, device s and methods used in acquiring data. Chapter III presents the testing locations and raw data associated to each location. Chapter IV presents the findings and results

PAGE 15

2 from each test. Chapter V presents future research that can be addressed . Chapter VI is the conclusion and key takeaways from this paper . Literature Review Previous research was conducted before administering tests to determine a specific methodology. Visual Geo Localization and Location Aware Image Understan ding Amir R oshan Zamir was reviewed to see how geo tags might be shown throughout experiments. location aware applications applying a geo tag to images. However, Zamir mentions that many of the procedures which use g eo tags as their input require a precise geo location, particularly in the urban areas. In respects to other software applications, this data is important to be accurate for the software to run properly . This was acknowledged and the use of different tools will be utilized when processing the data to validate one another. In addition, this paper was helpful to see the differ ent ways of displaying and picking test locations to conduct experiments at . Analysis o f errors in EXIF metadata on mobile devices Ana Lucila Sandoval Orozco and David Manuel Arenas Gonzalez. This piece was used as reference to see how EXIF metadata in photos might look like and what errors may arise from experiments . Orozco and Gonzalez mention that the area of image forensics analysis can be divided into two large branches: picture authentication and source authentication. Moreover, Orozco and Gonalez states if GPS tags are in place in metadata and display values from 0 to 1, th is indicates that the data has a high probability of being wrong. T his type of occurrence was not seen in all experiments conducted . However, was taken into account because some data did not capture GPS coordinates and did not display any information whats oever. Orozco and Gonzalez

PAGE 16

3 explain forensic techniques for image analysis, describe image metadata and how this is reviewed. Smartphone GPS accuracy study in an urban environment Merry was another work that contributed to how the administration of photos will be handled to capture data at different points of interest. Merry explains how phone positioning can hinder the results. Merry states , that at the collection of th e first point collected, the phones WIFI capability was disabled. Further, a fter the collection of the first point the WIFI was enable d and two minutes were allowed to pass before the second data point was collected. While administrating tests, it was conf irmed that the tester would take t his into consideration and to have devices have a certain time allotted for GPS to be acquired . objective of her study was to determine the accuracy of an iPhone 6s location under GPS only and WIFI only settin gs .

PAGE 17

4 II. METHODOLOGY Description of Materials ( D evices U sed) Devices were used in a series of two tests and eleven locations comparing either an Apple, Samsung or Garmin GPS unit. Figure 1 illustrates the type of device, release date, OS version, network tech, WLAN, Bluetooth and GPS satellites associated to each device. Details from each device is gathered by GSMarena.com , wh ich is a well known resource for device specifications . It is important to note the type of GPS associated to each device . This can be an indicator that results should vary between each test . For example, Apple devices use a GPS satellite system titled, Q ZSS, where Samsung utilizes the BDS GPS satellite. Figure 1 : Devices U sed Software Multiple types of software were used throughout test ing . Many were chosen to validate other tools used. The different type of software and online resources used were Pic2Map, Google Earth, SARTOPO and JPEGSnoop. Below will list each software with a brief description and their function throughout the experiments .

PAGE 18

5 Pic2Map Pic2Map is an online EXIF data viewer wi th GPS support which allows the user to locate and view your photos on Google maps. Throughout these experiments this tool was utilized for ease of displaying EXIF data within test photos . Also, this tool featured the ability to extract and view GPS coordinates associated to test photo s . Pic2Map can be accessed by https://pic2map.co m and from there a user can upload image files for EXIF viewing. Google Earth Google Earth is a computer program that renders a 3D rep resentation of Earth based primarily on satellite imagery. This tool was utilized to validate other tool s in the testing environment. Google Earth is a great tool for GPS coordinate plotting as it displays each point throughout its updated satellite maps . Google Earth version 7.3 was used. SARTOPO SARTOPO is a mapping and trip planning tool for the back country. This tool is widely used in search and rescue operations where it provides simple ease of GPS coordinate plotting . SARTOPO utilizes different typ e of maps ranging from Google satellite imagery to elevation maps. SARTOPO was a tool mainly used for displaying GPS coordinates and was compared to Google Earth for verification. SARTOPO is an online tool and can be accessed from https://sartopo.com/ . JPEGSnoop JPEGSnoop is a software that scans the image and offers the user all the detailed information called EXIF data . EXIF data contains information about the camera, edition program, date, color histogram, compression formats and other details associated to the image metadata. JPEGSnoop was used within these experiments as another tool to test data for

PAGE 19

6 accuracy and to provide ano ther way for viewing test photos. JPEGSnoop software version 1.7.3 was used. Matlab Matlab is a software that combines a desktop environment turned for iterative analysis and design processes with a programming language that expresses matrix and array mat hematics directly ( https://www.mathworks.com/products/matlab.html ). Matlab was used through these experiments to calculate the distance between actual known points against coordinates gathere d from the test photos. Matlab version 9.7.0.1190202 was used. GPS Explanation Understanding a basic idea of how GPS satellites talk to a device is important to grasp a sense of what kind of processes a device may be experiencing in the background. A brief utiliz e on E arth for acquiring GPS . GPS is a group of 24 or more satellites fl ying above the surface of Earth. E ach one circles the planet twice a day in one of six orbits to provide continuous, worldwide coverage. GPS satellites broadcast radio signals provid ing their locations, status, and precis e time from on board atomic clocks. GPS radio signals travel through space at the speed of light, more than 299,792 km/second. GPS devices on Earth receive the radio signals noting thei r exact time of arrival and use these to calculate its distance from each satellite in view. Once a GPS device knows its distance from at least four satellites, it can use geometry to determine its location on Earth in three dimensions ( https://www.gps.gov/multimedia/poster/ ) .

PAGE 20

7 The different satellite systems used throughout the tests are A GPS, GLONASS, GALILEO, QZSS and BDS. Below is a brief description of each satellite system associated to devices in this experiment that are liste d in Figure 1. A GPS A GPS (Assisted Global Position System) is a procedure GPS chips use to provide accurate positioning. Use of cell service, WIFI and latest GPS system available to provide an location as soon as possible to the device ( https://www.windowscentral.com/gps vs agps quick tutorial ) GLONASS GLONASS (Globalnaya Navigazionnaya Sputnikovya Sistema) is a global navigations satellite system owned and operated by the Russian Federations ( https://www.gps.gov/system s/gnss/ ) GALILEO GALILEO is a global navigations satellite system owned and operated by the European Union ( https://www.gps. gov/system s/gnss/ ) QZSS QZSS (Quasi Zenith Satellite System) is a global navigations satellite system owned by the Government of Japan and operated by the QZS System Service Inc. (QSS). QZSS complements GPS to improve coverage in East Asia and Oceania ( https://www.gps.gov/system s/gnss/ ) BDS BDS (BeiDou Navigation Satellite System) is a regional global navigations satellite system https://www.gps.gov/system s/gnss/ )

PAGE 21

8 EXIF Photo GPS Explanation from JPEGSnoop EXIF p hoto data is the metadata associated to the photo. Focusing on EXIF GPS data can provide quite a bit of details relating to the image file . Figure 2 is an example of a photo being uploaded to JPEGSnoop and the output referencing GPS EXIF data . JPEGSnoop reads the GPS data associated by its offset in h ex relating to the GPS L atitude Ref, GPS L atitude, GPS L ongitude Ref, GPS L ongitude, GPS A ltitude Ref, GPS A ltitude, GPS T imestamp, GPS P rocessing M ethod and GPS D ate S tamp. The alteration of this data cannot be done, as it would change the metadata and ultimately become a new image file . Below is a brief explanation of each type of result JPEGSnoop produces from the GPS EXIF data. coordinates are being captured from the device. GPS Latitude displays the latitude of the image file in degree and meters (varies on device and how GPS is being captured) coordinate are being captured from the device. GPS Longitude displays the longitude of the image file in degr ee and meters (varies on device and how GPS is being captured) GPS Altitude displays the elevation of the photo being captured usually in meters GPS Time Stamp displays the time by hours, minutes and seconds GPS Processing Method records the name of the method used for location finding ( https://www.exiv2.org/tags xmp exif.html )

PAGE 22

9 GPS Date Stamp displays the date of the image file being captured by Year, Month and date Figure 2 : EXIF GPS Example: JPEGSnoop Figure 3 displays the hex view of this information from the given offset seen in information provided by JPEGSnoop. This offset is the start of the EXIF GPSIFD for this image file. Even following the next parameter , GPSLatitudeRef, it can be shown in Figure 3 AC S II view that the latitude reference of N is shown. From results within tests administered, EXIF GPS headers did not display in an image file if that file did not capture any GPS coordinates. Figure 3 : EXIF Hex View (HxD) Method Used to Acquire Data In the test phase, two different ways of administering tests were done at eleven different locations . This was due to finding out new information and curiosities that arose from previous tests and locations . I nitial T est serie s #I, focused on the structure order of having cell service active , first photo taken, device switching to airplane mode then second photo taken . This

PAGE 23

10 process was working, however was later theorized that the devices could be saving information from previous locations and applying that information to newly created images. In turn of this curiosity , the administration of two more tests were done . Test ser ies #II was conducted to see if having airplane mode on initially would make a difference in results. Additionally, actual point locations between Test series #I was acquired by plotting the point on SARTOPO and making that point the baseline for each te st location . Test series #II , introduced the adoption of NGS survey markers , as the actual location parameter to use against the tested image files. Below are the steps used for the two different types of tests that were administered throughout this experi ment . Test #I Locations #I I X I. Both cellular devices and GPS units were placed at a stationary position at the The tester then waited for the GPS unit to display a GPS coordinate. II. Then a cellular device was used to take a picture of the GPS unit with cell service on (if cell service was available on the device). This triggers taking a picture of known GPS coordinates with t he device creating in the picture the test GPS coordinates. III. The cellular device was then put in airplane mode under the device setting s . Then a nother photo was taken. IV. The second GPS standalone device was handled and set in the same location . Steps #II III were repeated with the cellular device and the second GPS device. V. W ith a new cellular device , S teps # II IV were repeated VI. Tests concluded once all cellular devices foll owed steps #II IV

PAGE 24

11 Figure 4 : Photo Example from Cell Phone Device VII. Extraction of images from each cellular device was done by plugging each phone into a computer and extracting images to a USB device. This was to ensure best practices in preserving any metadata associated to the image file s . VIII. Image files were then examined using JPEGsnoop and Pic2map. EXIF data was extracted and organized in Microsoft Exce l. IX. Once all data points were extracted, each point was plotted by latitude and longitude locations from the EXIF data using SARTOPO. X. Once plotted , maps were extracted f rom SARTOPO to a .kml file format to view using Google Earth. Test #II Locations #X #XI: NGS Locations and Initial Airplane Mode Tests I. Cell ular device was put into a irplane m ode and then powered off. Cellular device was then powered back on ensuring airplane mode was still active. II. GPS unit was placed on top of NGS location marker an d tester waited for a GPS coordinate to be displayed on the GPS unit.

PAGE 25

12 III. Photo taken from cellular device of GPS unit on top of NGS survey marker. This triggers taking a picture of known GPS coordinates with the device creating in the picture the test GPS co ordinates IV. With a new cellular device, S teps # I III were administered. V. Extraction of images from each cellular device was done by plugging each cellular device into a computer and extracting images to a USB device. This was to ensure best practices in preserving any metadata associated to the image file s . VI. Image files were then examined using JPEGsnoop and Pic2map. EXIF data was extracted and organized in Microsoft Exce l. Analysis to D etermine Elevation Error Elevation e rror at each location was determi ned in two different ways. Test series #I, SARTOPO. This actual location elevation was compared to the elevation being displayed by each t est images EXIF data. With Test series #II , elevation was de termined by the known elevation listed in the NGS survey marker data sheets. The elevation listed in the data sheets were compared to the elevation being displayed by each test EXIF data. With having a known and test data values , we can determine the percent error associated to each image by using the percent error formula listed in Figure 5 . This formula was conducted in each test data set and a percent error was addressed for each test image file. After each test image file had a percent error as sociated to it, an average percent error was gathered for each device. These values are displayed in each test location data set.

PAGE 26

13 Figure 5 : Percent Error Formula Analysis to Determine Distance from Actual Location In order to calculate an error between GPS coordinates, the distance between two points was the best way to display this type of error. From having an actual location coordinate and coordinate from each photo we can determine the length in meters between b oth coordinates. This calculation was done using a Matlab script. This script is associated to the distance formula and Matlab defines the script as computing the lengths of the great circle arcs connecting pairs of points on the surface of a sphere, in ea ch case the shorter arc is assumed ( https://www.mathwork s. com/help/map/ref/ distance.html#d117e2 0321 ). Matlab Script: [arclen, az] = distance(lat1,lon1,lat2,lon2)*1000 The above script produced the number of meters from the actual location against the coordinates extracted from the experimental images. In each test, lat1 and lon were the same values and represented th e actual location point. They were measured which represented the latitude and longitude coordinates produced by each test image. In each test data set , all coordinates gathered were converted to decimal degrees GPS format. This degree format does not change the location of the original acquired coordinate. This conversion was necessary for Matlab to be able to process the distance difference. The distance from actual location from each image file and average actual location for each device test is displayed in each test data set in Chapter III.

PAGE 27

14 III. TESTING LOCATIONS AND RESULTS FROM TEST #I & #II All tests will display a raw data table pertaining to the image file results and list s columns by file name, the device, the GPS satellite that device uses, date, time displayed in GMT 07:00, GPS l atitude reference, latitude, GPS longitude reference and elevation. All tests latitude and longitude coordinates are redacted by only the in itial degrees portion of the coordinate. Color coordination is also implemented to show a constant color for each device being tested. This color coordination stays consistent with the image file pertinent to the device used to show a source of where inf ormation originated from for the Garmin device photos. For each test another table will follow that focuses on using the raw data to display what was analyzed . The analyzed data are in relation to the elevation percent error, average percent error of elev ation, the distance each image file from the actual point (meters) and the average distance from the actual point (meters) relating to the device. Test # I Rural Location # I Rural Location #I , used the iPhone 6s, iPhone 7 Plus, Samsung SM N960U, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 1. Below will list some observations listed in these data sets.

PAGE 28

15 Table 1 : Test # I Rural Location # I Coordinates and Elevation Table 2 reveals that the iPhone 6s , jpg . jpg location. By referencing Figure 6 the difference gap is shown . This is mainly due to the device trying to update its location to a much more precise one. This is worth noting because the device was in a irplane mode and possibly still trying to update its own location. Figure 6 : Rural Location # I Anomaly IMG_0119_Airplane.jpg L ocation C ompared to A ctual L ocation ( SARTOPO M ap )

PAGE 29

16 Consistent distance of approximate 12 meters 15 meters away from the actual location are acquired from the iPhone 6s, iPhone 7 Plus, Samsung SM N960U and the Garmin eTrex 20x. Garmin GPSmap 62s displays a location roughly 80 meters away and stays consistent with producing that result. Figure 7 and 8 , display SARTOPO and Google Earth maps displaying an overall view of all image file GPS coordinates plotted compared to the actual location. Figure 7 : Rural Location #I Image GPS C oordinates vs Actual Location (SARTOPO map) Figure 8 : Rural Location #I Image GPS C oordinates vs Actual Location (Google Earth map)

PAGE 30

17 As for elevation, the percent error across all devices is consistent pertaining to each device. A nomalies with the iPhone 6s were displayed initially with jpg jpg showing a n initial percent error of 61.81%. This is then adjusted to a .87% error j and stays consistent with the three image files that followed . It is worth noting that the Samsung SM N960U elevation error was the worse with an approximate elevation percent error of 9% 10%. Table 2 : Rural Location # I Percent Error Elevations and Average Distance from Actual Test # I Rural Location # II Rural Location #II used the iPhone 6s, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. It is worth noting that a Samsung device was not utilized in this test. The raw data associated to this test is displayed in Table 3. Below will list some observations listed in these data set s .

PAGE 31

18 Table 3 : Rural Location # II Coordinates and Elevation Table 4 and Figure 9 display an anomaly taking place. Starting with the iPhone 6s and jpg . T here is a change of distance from actual, starting from a close distance and exceeding to one that is more than double. B etween th ese two image files , airplane mode is being turned on and producing similar coordinates to file jpg might be the cause for the sudden increase of distance possibly having the device relying on other connections . Figure 9 : Rural Location #II Image GPS C oordinates vs Actual Location (SARTOPO map)

PAGE 32

19 Looking at the iPhone 7 Plus and at jpg an initial distance of 21,717.1 meters away from the actual location . With in 13 seconds after that image file was jpg eters away from the actual location . This brings out another observation being made with the switch to airplane mode . Continuing with the same device the GPS coordinates from images produced after the previous image are displayed and narrowing in on the actual location where the device is present (Figure 1 1 ) . Figure 10 : Rural Location #II Anomaly IMG_1950.jpg D istance from Actual Location (SARTOPO map) Figure 11 : Rural Location #II Anomaly #2 iPhone 7 Plus D istance C orrections to Actual Location (SARTOPO M ap)

PAGE 33

20 Garmin eTrex 20x was stable in respect to its distance data points . T he Garmin GPSmap62s had distances displaying approximately 77 meters away from the actual location and possessed jumps to approximately 100 meters. Elevation was consistent . jpg . T he image fil e produced an elevation of 5 meters, which in turn produced a 99.37% error from the actual elevation. ne. jpg Again, another trend with airplane mode change creating a more accurate result. Table 4 : Rural Location #II Percent Error Elevations and Average Distance from Actual Test # I Rural Location # III Rural Location #III used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 5. Below will list some observations listed in these data sets.

PAGE 34

21 Table 5 : Rural Location # III Coordinates and Elevation From referencing Table 6, the iPhone 6s had the wors e cellular device average distance from actual location . The Samsung Galaxy S6 had no error in relation to distance . T his device w as administered s trictly on airplane mode and did not have cell service active . With this stipulation the Samsung Galaxy S6 took roughly 30 secs to 5 mins to acquire a GPS coordinate. This eludes to the device being able to produce a more than accurate coordinate. Mainly considering the coordinate being produced is being solely rel iant on the GPS chip on the device . The iPhone 7 Plus and Garmin eTrex 20x had variations between its distance, but nothing too alarming. Garmin GPSmap 62s again produced far distance s from actual location results staying within approximately 80 meters 9 1 meters. Figure 1 2 displays a SARTOPO map of the image GPS coordinates compared to actual location.

PAGE 35

22 Figure 12 : Rural Location #III Image GPS C oordinates vs Actual Location (SARTOPO M ap) Elevation showed the iPhone 6s had the wors e average percent error and this might share a correlation with that phone also producing poor distance results within this test . The other devices produced elevation errors that were not worth addressing . Table 6 : Rural Location # III Percent Error Elevation and Average Distance from Actual

PAGE 36

23 Test # I Rural Location # IV Rural Location #IV used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 7. Below will list some observations listed in these data sets. Table 7 : Rural Location # IV Rural # IV Coordinates and Elevation Referencing Table 8 and Figure 1 3 , it is show n th at the iPhone 7 Plus displays the wors e average distance from actual location for a cellular device. The Samsung Galaxy S6 again displayed no error regarding distance. The iPhone 6s and the Garmin eTrex 20x produced a semi constant result. GPSmap 62s again stayed within its approximately 80 meter distance error. Figure 13 : Rural Location #IV Image GPS C oordinates vs Actual Location (SARTOPO M ap)

PAGE 37

24 Figure 14 : Rural Location #IV Image GPS C oordinates vs Actual Location (Google Earth M ap) Figure 1 4 displays a Google Earth map to display a visual of the type of environment where this test was conducted . From data in Table 8, the iPhone 7 Plus had the wors e average percent error. This again eludes to the elevation and distance error shar ing the same error correlation. However, w hat goes against this correlation is the Samsung Galaxy S6 . The Galaxy S6 produced no error whatsoever with distance but displayed a 15.35% error in elevation. This further can elude to this type of trend bein g noticed based on the devices itself, rather than the distance and elevation error sharing the same type of error rate .

PAGE 38

25 Table 8 : Rural Location # IV Percent Error Elevation and Average Distance from Actual Test # I Suburb Location # I Suburb Location #I was used the iPhone 6s, Samsung SM N960U, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 9. Below will list some observations listed in these data sets. Table 9 : Suburb Location # I Coordinates and Elevation In referencing Table 10, the biggest anomaly that stands out is image file N960U. This image file was 15,164.6 meters from the actual location and a 90.60% error rate regarding elevation. It is interesting that this image file was the l ast image taken during the test . T he SM N0 60U previously captured image files

PAGE 39

26 producing a 18.3948 meter distance from actual and a 42.59% elevation error rate. The cellular device was placed in a irplane mode . H owever, this could be something going on in the background in the device priority list o f how it acquires location and was trying to correct itself. Figure 15 displays the distance between the two points. Figure 15 : Suburb Location #I Anomaly Samsung Note 9 20190916_102914_Airplane.JPG D istance from Actual Location (SARTOPO M ap) Another anomaly was with t he iPhone 6s . This initially produced an approximate 30 meter distance from actual location , but then corrected to an approximate 15 meter from actual point. The other devices stayed consistent with the trends being associated to them in previous tests. Figures 16 and 17 display s the test image GPS coordinates vs actual location. Figure 16 : Suburb Location #I Image GPS C oordinates vs Actual Location (SARTOPO M ap)

PAGE 40

27 Figure 17 : Suburb Location #I Image GPS C oordinates vs Actual Location (Google Earth M ap) Table 10 : Suburb Location # I Percent Error and Average Distance from Actual Test # I Suburb Location # II Suburb Location #II used the Samsung SM N960U, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 11. Below will list some observations listed in these data sets.

PAGE 41

28 Table 11 : Suburb Location # II Coordinates and Elevation In reference to Table 12, the iPhone 7 Plus had a slight fluctuation from close distances to actual location to farther ones . This is not so alarming seeing this type of movement due to the device constantly trying to adjust its location to provide a preci se location. It is worth noting that the Samsung device provided the worse average distance from actual location and the worse elevation percent error compared to the iPhone 7 Plus. The Garmin eTrex provided the best average distance from actual point . Figures 1 8 and 1 19 display SARTOPO and Google Earth maps displaying distances between each device image file coordinates. Figure 18 : Suburb Location #II Image GPS C oordinates vs Actual Location (SARTOPO M ap)

PAGE 42

29 Figure 19 : Suburb Location #II Image GPS C oordinates vs Actual Location (Google Earth M ap) With elevation the iPhone 7 Plus provided the best percent error of 1.57% compared to all devices . The Samsung device and Garmin units surprising provided relatively high percent errors compared to the iPhone 7 Plus, ranging from 22.50% to 40.94%. Table 12 : Test # VI Suburb # II Percent Error Elevation and Average Distance from Actual Test # I Suburb Location # III Suburb Location #III used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 13. Below will list some observations listed in these data sets.

PAGE 43

30 Table 13 : Suburb Location #III Coordinates and Elevation Table 14 shows the Samsung Galaxy S6 possessed no error rate regarding distance but displayed a 37.66% error rate regarding elevation. T he iPhone 6s and iPhone 7 Plus average distances from actual location did not produce alarming results . T hey were consistent and close to the actual location with 8 meter to 13 meter differences . Garmin eTrex 20x had a better average distance from actual location than the iPhone 7 Plus and iPhone 6s . Garmin GPSmap 62s still produced poor results by have the worse av erage distance from actual location. Figure 20 displays the SARTOPO map of all image file GPS coordinates compared to actual location.

PAGE 44

31 Figure 20 : Suburb Location #III Image GPS C oordinates vs Actual Location (SARTOPO M ap) With elevation, the iPhone 6s and iPhone 7 Plus had a different elevation profile that was not too similar. The iPhone 6s produced an average elevation percent error of 7.26% and the iPhone 7 Plus produced an ave rage percent error of 3.23%. From these devices being only a year device generation apart, it was interesting to see that these error rates will not be relatively closer. The Garmin eTrex 20x did not produce a good elevation profile in comparison to the c ellular devices. Garmin GPSmap 62s again produced poor results by having the worse average elevation percent error. Table 14 : Suburb Location # III Percent Error Elevation and Distance from Actual

PAGE 45

32 Test # I Suburb Location # IV Suburb Location # I V used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 15. Below will list some observations listed in these data sets. Table 15 : Suburb Location #IV Coordinates and Elevation Table 16 displayed the Samsung Galaxy S6 producing the worse average distance from actual location with in cellular device s and the worse average elevation percent error categories . The two image files that were produced by the Galaxy S6 were also taken with 10 seconds apart and did not produce different results. The iPhone 7 Plus outperformed all devices in distance and the GPSmap 62s outperformed all devices in elevation. Figure 21 displays an overall view of all image GPS coordinates vs the actual location.

PAGE 46

33 Figure 21 : Suburb Location #IV GPS C oordinates vs Actual Location (SARTOPO M ap) There was one distinct change in distance with one device showing the device correcting itself closer to the actual location. T jpg file produced a distance of 6.717 meters from actual location. P revious images associated to this device produced an approximate 17 meter 24 meter distance from actual location. The amount of time the device took for this correction was 40 seconds. Table 16 : Suburb Location # IV Percent Error and Average Distance from Actual

PAGE 47

34 Test # I Urban Location # I Urban Location #I used the iPhone 7, iPhone 8, iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. The raw data associated to this test is displayed in Table 17. Below will list some observations listed in these data sets. T able 17 : Urban Location # I Coordinates and Elevation From looking at results from Table 18, iPhone 7 Plus had the worse results of average jpg jpg the initial image file produced a 13.5465 meter distance from actual location . T hen an image taken 1 min and 16 seconds later produced a 208.5322 meter distance from actual location . This was the biggest jump seen in this test. However, the iPhone 7 produced jpg jpg from the iPhone 7 Plus, where the image file coordinate was first providing a far distance but then corrected to a closer distance.

PAGE 48

35 It was of interest to see the iPhone 8, iPhone 6s, Samsung Galaxy s6 and the Garmin eTrex 20x were able to provide a distance stable to one another . Samsung Galaxy S6 was able to produce no error in distance again, which was surprising with being in an urban environment. However, its elevation percent error was 100%, in which the device was not able to produce an elevation whatsoever . Figures 22 an d 23 display SARTOPO and Google Earth maps displays image GPS coordinates compared to actual location. Figure 22 : Urban Location #I Image GPS C oordinates vs Actual Location (SARTOPO M ap) Figure 23 : Urban Location #I Image GPS C oordinates vs Actual Location (Google Earth M ap)

PAGE 49

36 Moving to overall elevation in this test , the Garmin GPSmap 62s produced the worse average elevation percent error rate at a 363.72%. The worse cellular device elevation percent error was the iPhone 7 at a 132.71%. The iPhone 8 was able to produce the best elevation average percent error profile at a 3.81%. Table 18 : Urban Location # I Percent Error Elevation and Average Distance from Actual Test # II NGS Location #I Urban #II NGS Location #I Urban #II, used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. This was the first test that utilized the NGS survey markers. The survey marker was in downtown Portland, Oregon , USA . Figure 24 displays a photo of the NGS survey ma ker with the NGS data sheet that displays the GPS coordinates and the elevation of the marker. This test was the initial implementation of putting our devices on airplane mode, powering off device, powering on device ensuring airplane mode is still active and capturing the initial photo in the test series . The raw data associated to this test is displayed in Table 19. Below will list some observations listed in these data sets.

PAGE 50

37 Figure 24 : NGS Location #I Urban #II NGS Data Sheet Rural Environment and Survey marker 54 RESET Table 19 : NGS Location #I Urban #II Coordinates and Elevation Table 20 shows the iPhone 7 Plus with both the worse average distance from actual location and average elevation percent error. This could be due to the fact of this device initially Aiplane. jpg _ 2014 Airplane. jpg these initial photos did not obtain a GPS location or elevation. This is due to the devices c takes approximately 25 secs for the device to acquire any GPS location. I believe this sam e result Airplane. jpg file . However, its next image file was

PAGE 51

38 able to provide a pretty accurate distance from actual location . Figure 2 5 displays the image GPS coordinates compared to the actual GIS survey marker locat ion. Figure 25 : NGS Location #I Urban #II Image GPS C oordinates from Actual Location (SARTOPO M ap) It is worth noting that the Samsung Galaxy S6 was not able to provide an elevation profile . H owever, did provide a location point of 29.8002 meters from actual location . T he stand alone GPS units produced similar results as previous tests administered . Table 20 : NGS Location # I Urban # II Percent Error and Average Distance from Actual Test # II NGS Location #II Rural # V NGS Location #II Rural #IV, used the iPhone 6s, Samsung Galaxy S6, iPhone 7 Plus, Garmin eTrex 20x and the Garmin GPSmap62s. This was the second test that utilized the NGS

PAGE 52

39 survey markers. The survey marker was at a rural site used in a couple previous tests in our rural environment series outside of Portland, Oregon , USA . Worth noting , this test was a couple meters away from the previous actual location . Figure 2 6 displays a photo of the NGS survey maker with the NGS data sheet that displays the GPS coordinates and the elevation of the marker. This test also implemented our second test putting our devices on airplane mode first, powering off device, powering on dev ice and capturing a photo. The raw data associated to this test is displayed in Table 21. Below will list some observations listed in these data sets. Figure 26 : NGS Location #II Urban #V NGS D ata S heet Urban Environment and Survey Marker RD 2197

PAGE 53

40 Table 21 : NGS Location #II Rural #IV Coordinates and Elevation From Table 22, it is displayed that mostly all devices provided an elevation and GPS profile. It was discovered for this test that the iPhone Camera App option for location needed to ask the user if location is desirable for the app session. This seems to enable GPS right away when enabled, as the device was mostly receiving GPS coordinates right after exiting that prompt. The only anomaly with this location setting was from t Airplane. jpg However, a GPS Airplane. jpg Airplane. jpg time offset was not listed like the other photos EXIF data .

PAGE 54

41 Within this location test the GPSmap 62s provided the worse average distance from actual location followed by the Samsung Galaxy S6, which provided the worse distance for a cellular device. good results in the distance category. As the iPhone 6s received an average distance of 3.1061 meters from actual location and th e iPhone 7 Plus received an average distance of 2.36365 meters from actual location. The Garmin eTrex 20x provided the best average distance of 1.5547 meters from actual location. Reference Figure 2 7 for all image GPS coordinates compared to the actual GPS survey marker location. Figure 27 : NGS Location #II Urban #V Image GPS C oordinates vs Actual Location (SARTOPO Map )

PAGE 55

42 Figure 28 : NGS Location #II Urban #V Image GPS C oordinates vs Actual Location (Google Earth M ap) With elevation the Samsung Galaxy S6 provided the best average elevation percent error of 0.44%. Surprisingly the Garmin eTrex 20x provided the worse elevation percent error of 2.31%. It was interesting to see that all device s did well in both elevation and distance compared to previous data sets. Figure 27 displays a Google Earth map to show the type of elevation associated to the environment type. Table 22 : NGS Location #II Urban #V Percent Error Elevation and Average Distance from Actual

PAGE 56

43 IV. FINDINGS AND RESULTS EXPLANATION Below will describe findings relating to distance and elevation error depending on device and environment type. The findings between Apple and Samsung device will the discussed. Observations with airplane and other notable findings will be addr essed. Distance Error By a cquiring the average distance in meters from each photo taken against the known actual location using the Matlab script : [arclen, az] = distance(lat1,lon1,lat2,lon2)*1000 . Figure 2 9 displays each test location in respect to each device that was used. An overall average distance was calculated to display another figure of accuracy between all tests associated to the device. Additionally, an overall average distance for each environment type per each device was calculated and d isplayed to show which device excelled in different environment types. Below will explain what was observed in the device and environment types regarding distance. Figure 29 : Distance Error from Actual Location

PAGE 57

44 Device From data in Figure 2 9 , it can be shown that the Samsung Galaxy S6, iPhone 8 and the Garmin Etrex 20x displayed an average distance closest to the actual location (0). However, the iPhone 8 was only used in one experiment, so it cannot be proven that it displayed a better GPS accuracy. The next runner up would be the Samsung Galaxy S6, which makes sense merely that on some tests the phone displayed no distance error. The worse devices displayed from Figure 2 9 were the Samsung SM N960U and the iPhone 7 Plus. These were most likely due to anomalies of GPS locations putting the devices in very far out places when trying to secure a promising GPS coordinate. T wo instances , one from each device showed this anomaly and will be explained in the other findings section later throughout the paper . However, I believe these two outliers contributed to these devices performing poorly by providing the overall worse average distance from actual location . Environment Figure 30 : Device Average Distance P ertaining to Environment From Figure 30 , Garmin devices appear to have the most accurate distances. Apple being second closest and Samsung being the farthest away from the actual. However, these differ from

PAGE 58

45 each type of environment. This figure also paints a picture of how these environment typ es vary as far as accuracy all together. Looking the overall rural figure for both companies. overall rural figure is 742.1786 meters from actual location and actual location. But in a s uburb e nvironment these two companies both switched roles . Samsung suburb figure to 16.1289 meters from actual location. Below will outline each environment type for the best an d worse devices for each setting. Rural From Figure 30 and looking at overall rural results . Samsung devices appear to be closest to the actual location, with Garmin GPS units following and Apple devices being the farthest from actual location. Apple and Samsung devices collectively have a rural overall average distance of 377.3447192 meters. All devices in this environment have an average distance of 225.7768765 meters. Figure 3 1 displays a scatter plot of al l rural location tests compared to the actual location point. the latitude point in this graph is the major defining factor of where that point would lie in relation to the actual point plot.

PAGE 59

46 Figure 31 : Rural Distance I mage GPS P lot vs Actual Location Plot Scatterplot Graph Suburb From Figure 30 and looking at overall suburb results. Apple devices appear to be closest to the actual location, with Garmin GPS units following and Samsung devices being the farthest from actual location. Apple and Samsung devices collectively have a suburb overall average distance of 491.1100914 meters. A ll devices in this environment have an average distance of 201.0655325 meters. Figure 3 2 displays a scatter plot of all suburb location tests compared to the actual location point. Again a trend is displayed of the longitude staying consistent for the act ual point

PAGE 60

47 Figure 32 : Suburb Distance Image GPS P lot vs Actual Location Plot Scatterplot Graph Urban From Figure 30 and looking at overall urban results. Samsung devices appear to be closest to the actual location, with Apple devices following and Garmin GPS units being the farthest from actual location. Apple and Samsung devices collectively have an urban overall aver age distance of 29.22586 meters. All devices in this environment have an average distance of 57.3394 meters. Figure 3 3 displays a scatter plot of the urban location test compared to the actual location point. This provides a nice visual of the inconsisten cies that arose from the different devices and what trends were demonstrated. From previous scatter plots graph the same trend that devices seem to follow are shown.

PAGE 61

48 Figure 3 3 : Urban Distance Image GPS P lot vs Actual Location Plot Scatterplot Graph Elevation Error Elevation Error was determined by using the percent error formula displayed in Figure 4. This was used to come up with a percent error for each elevation data point f rom each test image. An average percent error from each device test was plotted in Figure 3 4 . A n overall percent error for each environment was calculated and displayed in Figure 3 4 .

PAGE 62

49 Figure 34 : Overall Elevation Percent Error Device Figure 35 ranks each device from overall lowest average elevation percent error . Figure 35 : Elevation Average Percent Error Device Ranking Both the iPhone 8 and iPhone 7 have a red line through their results, due to both devices only being tested in one experiment. These devices cannot b e shown as either the best or worse device in respect to average percent error for all tests and will not b e taken account for the ranking . Glancing over each device shows that there is no constant percent error rate relating to each environment. There is much of a change depending on the device itself and not solely on any environment type .

PAGE 63

50 Environment From Figure 3 4 , each environment had different devices that excelled over others. Within this section each environments percent error for devices will be addressed. In a rural setting the Garmin eTrex had the lowest percent error. For cellular devices, the iPhone 6s had the lowest percent error. The worse device in this category was the iPhone 7 Plus. Having a Garmin unit providing the lowest percent error rate in this test was not too surprising as this is a stand alone GPS unit and its main f unction is to be able to provide ultimate accuracy in this type of environment setting. In a suburb setting the iPhone 7 Plus had the lowest percent error. The worse device in this category was the Samsung Galaxy S6. I t was interesting that the Apple devic es did not produce similar results overall. Since they are pretty on par in respects to the device models not having a huge generation gap difference. In an urban setting the iPhone 8 had the lowest percent error but was only used in one test. This leads t o the runner up, the iPhone 6s having the lowest percent error. The worse device in this category was the Garmin GPSmap 62s. It was intriguing to see t hat in a urban environment the Garmin devices did not excel, which might be due to other interference of devices and buildings that encompass an urban setting . Apple vs Samsung Apple and Samsung have varied results when compared to overall average distance from an actual location and different environments. Figure 3 6 displays each environment with an overall distance from an actual location . Apple devices are favored overall in accuracy from the

PAGE 64

51 tests in acquiring an average distance from an actual location and overall in a suburb environment. Where Samsung devices ar e favored in a rural and urban environment. Figure 36 : Apple vs Samsung Distance from Actual This is an interesting observation since these two companies use 3 similar satellites, however different in one. Maybe the BDS satellite system that Samsung utilizes favors rural and systems favoring suburb en vironments over rural and urban. It would make sense that these results would be different because one satellite system, in theory, should mean different results between these two companies. Airplane Mode From Test series # I , each cellular device followed the procedure of cell service on , then switching to airplane mode , then a photo being captured . However, it was then realized that the cellular device might be keeping a known location within the cache of the phone. This suggests after having the device change to airplane mode , the device may rely on the previous known location . Which in theory, the previous known location would be attached to the metadata of any new images. Test series # II were tests focused on how airplane mode alters GPS metadata to photos and if tests can prove that a cellular device might retain known locations from previous ph otos .

PAGE 65

52 From referencing section, Tests series #II : NGS Locations and I nitial A irplane M ode T ests, the phones were first put into airplane mode, powered off and then turned back on ensuring airplane mode was still active . From tests administer ed , t his revealed the phone does still capture GPS metadata, depending on device location settings . For Apple Devices, a user has three options to choose from under location services for captu ring locations fr om to allow location access to the camera app by either never , ask next time or while using the app . Under NGS Location #I , the iPhone 6s and iPhone 7 Plus did not initially provide a photo GPS coordinate, but the next photo did provide a GPS coordinate takes roughly 30 secs for the Camera app to trigger the GPS in the device to acquire a location for the photo. However, in NGS Location #II , . T his option triggers the c amera app to regularly turn on GPS right away displaying GPS coordinates in photo metadata. For Samsung devices, three opt ions are available to choose under location setting s . These three options are high accuracy , battery saving and device only . Under Test series # II , the and made the test solely based on the devices GPS . This process did take about five min utes for the device to acquire a GPS coordinate. However, a GPS coordinate was p roduce d within p hoto metadata . Other Notable Findings Random anomalies would occur with some of the GPS metadata associated to the photos. Within this section a discussion of notable anomalies will be addressed.

PAGE 66

53 One in particular was in Test # I Rural Location #II, with the iPhone 7 Plus. The test environment was rural and the first photo on that device displayed a coordinate that was 21717.1 meters away from the actual location. This could have been from the device taking some time to catch up to acquire a GPS satellite. However, most of the time this type of encounter that was this far of a distance in tests administered were not seen . Anothe r anomaly had to do with Test #I Suburb Location #I, with a Samsung SM N960U device. The interest in this anomaly shows a total of four photos taken within this test series. T he first three photos were near the actual location of the test series . However, the last photo brought the GPS coordinate all the way to a previous test location that was 15,164.6 meters away. Speculation think s that maybe the phone lost connection to a previous GPS satellite and only fell back to a previous known coordinate . Further cellular device forensics would have to be conducted to determine any other background processes causing this change. Previously noted before in this paper , the Samsung Galaxy S6 with no cell service activated will take roughly 45 secs to 5 mins to acquire a GPS coordinate. However, that coordinate most of the time was accurate to the actual location being tested. My assumption is that other connection types i.e. cell service, WIFI and Bluetooth could in fact hinder the device to determi ne a precise location. It was assumed that a device also utilizing these other connections would provide accuracy to the device . This could differ for each device model or software update and will need further testing. During tests, a situation came up when a photo taken in a rural location was displaying GPS coordinates of a residential area. implemented, which means that it will acquire a location from all connection types available. After some research of the GPS coordinate, it was determined that the device pull ed GPS

PAGE 67

54 coordinates from the connection to a WIFI router. Further cellular forensics would have to be done to explain any background processes occurring. But it is strange for this the cellular device to provide a location in this fashion.

PAGE 68

55 V. FUTURE RESEARCH Much more future research could be conducted on this topic focusing many different types of detail s photo EXIF data exhibits . Analysis of each device and operating system software versions from each cellular company could be a study to see if one operating system differs from another. It is possible for device architecture or software updates to be different and could change the priority list of how a devic e is determining a location. Also, a software update could maybe implement a new technology for the device to utilize connections more efficiently. Types of weather tests could further be done to determine if this causes GPS interferences . Maybe there can be a trend identified to determine any error offsets that could be done to account for this type of circumstance . Tests being catered around the anomalies that were being experience d throughout the paper would be another good research study. Mainly around what specific parameters are taking place in different devices and to explain how a device will output a certain GPS coordinate . It would probably be clear to have a forensic download of each cellular dev ice and analyze what processes the phone is going through when an anomaly occurs. B eing able to pinpoint what GPS satellites or satellite systems a device is talking to would be interesting to see if a device is for certain connecting to the closest satel lites . This eludes from previous statements focusing on further research between system system . Diving deeper into tests involving airplane mode would be another option. It was touched briefly with the last two tests administered and could be expanded to acquire more of overall

PAGE 69

56 determination of what processes may be happening in the background . Test on solely WIFI GPS coordinates would be beneficial to see the accuracy of device is a factor to known WIFI router locations. With the implementation of 5G this could become a trend in GPS accuracy.

PAGE 70

57 VI. CONCLUSION This study came forth with the intention to determine how accurate photo GPS data is from cellular devices compare d to GPS coordinates from a standalone GPS unit in different environments. It was administered in two different ways to de termine location accuracy , utilize calibrated known location coordinates and test airplane mode possible restrictions to provide a valid data set. Distance errors, elevation errors and anomalies were addressed . W hen these errors were compared to devices to see what trends or accuracy could be revealed. It was mainly addressed that device accuracies depended on different factors . The type of environment, cell service and other functions that may be going on in the background of the device could contribute to the cause of these anomalies . I t cannot be said if one device is more accurate than the other for this reason. Satellite systems that Apple and Samsung use within their device architecture were addressed and provided results that displayed different GPS coordinates and elevation profil es. S uggest ing that Apple and Samsung satellite choice may excel in different environment types and each device could be useful in different ways depending on the environment. Airplane mode was experimented in administering tests . The first series of tests had devices with ce ll service active first then switche d over to airplane mode . The second series of tests had the device powered off with airplane mode and powered back on with airplane mode active . It was found that depending on the device location settings , determines how fast the device may acquire a GPS c oordinate. Both series of tests types and still produced results that led to a GPS coordinate being created . However, t here is still a chance that after a device has been powered on

PAGE 71

58 and initial photos taken. There w ill not be a GPS coordinate produced depe nding on location setting s used . Environment tests were administered and does play a role in devices being able to acq uire an accurate GPS coordinate and elevation profile. Throughout tests it was displayed that devices favored an urban environment for GPS accuracy and least favored a rural environment. Many anomalies were involved in each test and seemed each device exceled in certain environments differently than others. Ultimately, many considerations come into play with GPS accuracy on smartphones. From research, it should be considered that claims made from GPS data in image files should be examined carefully . A ll factors of cell service, WIFI connection , environment an d device models play a role of an overall GPS coordinate created from a device . A GPS coordinate could be accurate or can have a random anomaly applied to it due to one of these factors. It has been displayed that different outputs of GPS coordinates are p roduced by different devices . Additionally , these outputs can be produce d by simply having a cel lular device in airplane mode. D ifferent devices excel in different types of environments relating to rural, suburb and urban areas . P hoto GPS data can be ver y beneficial in an investigation, especially if other forensic practices or key information of an investigation support any findings from GPS image data. F rom different anomalies previously addressed , having a validation standard implemented would be benef icial for the digital forensics community to have best practices set into place for use of EXIF image data. Even having a clearinghouse of cellular devices being tested against current GPS satellites would be a benefit for the digital forensic community . T his type of GPS data

PAGE 72

59 within images will become even more popular with new technologies being implemented in our way of life.

PAGE 73

60 REFERENCES Exiv 2. Metadata Reference Tables . n.d. 15 Oct 2019. . Explanation, Garmin GPS. https://www8.garmin.com/aboutGPS/ . 2017. 15 September 2019. Garmin. Garmin ETREX®10/20/20x/30/30x Owners Manual . Kansas: Garmin ltd, 2019. Garmin . Garmin Manual GPSmap 62s . Kansas: Garmin ltd, 2010 2011. gps.gov. https://www.gps.gov/systems/gnss/ . 18 December 2017. 15 September 2019. Merry, Krista, and Pete Bettinger. "SmartPhone GPS accuracy study in an urban environment." PloSone Vol 14,7 (2019). . National Geodetic Survey. Survey Maks and Datasheets . n.d. 16 October 2019. . Sandoval Orozco, A.L., Arenas González, D.M., García Villalba. "Analysis of errors in exif metadata on mobile devices." Mul timed Tools Appl (2015). . Zamir, Amir Roshan . " Visual Geo Localization and Location Aware Image Understanding ." University of Central Florida (2014). .