[OSINT/GEOINT] Using shadows and optics to geolocate a photo in a US military base
While browsing the internet, I came across this photo, it was accompanied by the following legend: Summer 2019, Middle East. Additional context was present with the photo (name of the military base), but it was deliberately not taken into account, in order to have a more complete investigation. I tried to do my best not to let this distort my investigation.
The type of location and his localisation intrigued me, and since I never had done geolocation in the Middle East, so I thought this would be a good exercise.
This investigation allowed me to establish a research methodology when we have access to the metadata of the photo (date, phone model).
A US military base in the Middle East?
The first step was to identify the type of place it could be and in which geographical area it might be located.
According to the caption, the photo was taken in the Middle East. This can be partially confirmed, as the stop sign (yellow) is in Arabic.
In the background we see containers (purple) and in the foreground barbed wire on the fence (red). There also seems to be a rescue truck behind the fence (“RESCUE”?).
In addition, the street names (green) correspond to the names of US cities and states. This could mean that it is an American military base.
With the first elements that we have recovered the hypothesis is that the photo was taken in a US military base in the Middle East. But which one?
Identification of the military base
The power of shadows
The next step was to identify which military base it is. To do this, I started by retrieving a list of US military bases in the Middle East from the American Security Project.
We could wander around on the maps of all the bases to find the exact location, but that would probably be too long and we might miss it if we are not methodical enough. We can however use a specificity of the base visible on the photo and search only for this specificity.
To do this, we can use the photo metadata to our advantage. Indeed, in this case, they contain the date on which the photo was taken.
$ exiftool image.jpg
[...]
Date/Time Original: 2019:09:12 13:29:35
[...]The date and time are interesting because we have a shadow on the photo. This one tells us that the sun is nearly aligned with the “Virginia Ave” sign. The sign being parallel to the road we are on, it means that if we can determine the position of the sun we can deduce the orientation of the road.
Even if you don’t have metadata, you can try to use the context of the photo and the shadows to deduce the angle of the road, as Nick Waters was able to do when investigating strikes in Gaza.
We can then use SunCalc to identify the position of the sun when the photo was taken. We just have to be careful with the timezone, because the time in the metadata is the phone time (local time) and not the UTC-0 time. We use the city of Riyadh as a reference point.
We can now measure the angle between the orientation of the sun and a horizontal line when the photo was taken. In our case this angle almost corresponds to the angle of the road where the photographer was.
The angle is 36°, but the sun is not perfectly aligned with the sign, we must remove 2°.
We now know that the road on which the photo was taken forms an angle of 34° with a horizontal line. We can now generate an image with the orientation of the road, in order to compare it with satellite images of US military bases in the Middle East.
We use the OnTopReplica tool to overlay the road orientation with transparency over the map. This allows us to identify six possible bases.
Street naming convention
Some of the bases can be quickly removed from the list, either because the environment or the language do not match. The last 3 are left, to differentiate them we can use the street names, using Google Maps and OpenStreetMap.
Camp Arifjan uses a naming convention similar to the streets in the photo. It is the only base in the list to do so. Unfortunately Washington Road and Virginia Ave are not known.
Location of the street
Now that we have the base, we must identify the place where the photo was taken. The base is quite large and has many roads corresponding to the angle we have identified. To simplify the visual search we can draw a more complete top view that we will overlay on the map.
A little bit of optics
In order to draw a top view, we need to identify the distances between the elements present in the photo.
The two small white buildings (purple) and the cellular tower (green) may be interesting to include in the top view, as well as the road along the containers and its intersection with “Washington Road”.
In order to calculate the distance between the camera and an object we need the focal length of the camera and the sensor height. This data can be indirectly retrieved from the metadata.
$ exiftool image.jpg
[...]
Camera Model Name : Moto G (4)
[...]Knowing the phone model we can retrieve its characteristics from the Camera FV-5 Device database:
- Focal length: 3.6mm
- Sensor height: 3.5mm
Before we can calculate the distance to an object we need to know its real size. For the two buildings we can compare them to the near containers, a container is 2.59m high, so we estimate a height of 3m.
(3.6*3000*714)/(16*3.5) = 138m
We are missing only two values, if we can identify one of them we can calculate the other because we are in a right triangle.
Based on the containers we can extrapolate and estimate the distance between the two buildings and “Washington Road”.
The buildings seem to be in front of the gray/green container.
And we can estimate the distance between the crossing and the grey/green container to be the equivalent of 10 containers, or 120m.
Using the Pythagorean theorem, we can now calculate the last side of the triangle: 68m.
We can now modify the image to have the right proportions.
It only remains to orient it.
The final step
Before you can use the top view on Google Earth using OnTopReplica, we need to calibrate it with the ruler tool so that it is the right size on the map.
In order not to miss the location we separate the map into several areas in Google Earth that we process as we go along.
And we finally find the right place!
Conclusion
We can then validate the location with the following elements:
- The wall (yellow)
- The buildings (purple)
- The cellular towers (green)
- The bushes (blue)
- The containers (red)
https://www.google.com/maps/place/28.88107055639882+48.15729104462818
To conclude, despite the different assumptions made when creating the top view, the proportions were not very far from reality, but there was a difference of several dozen meters. However, this made it easier to find the location of the photo. The area to be searched was relatively small (4.5km x 2.5km) and could have been done without a top view. It was mainly a pretext for me to implement this methodology, which could be useful when searching on larger areas.
Hoping that this methodology will help you in your investigations!
