To analyse shadows in source imagery, Shadow Finder maps all points on the earth where a shadow of given length could occur at a given date & time, IF the height of the object casting it is known.
https://github.com/bellingcat/ShadowFinder
Enter shadow length and object height, (or, instead, the angle of elevation of the sun) for a research source image, then also the date and time of the image, and see the potential locations at which such a shadow could be created for that point in time as the bright area (circle) on the world map.
The bright yellow area of the circle marks all the locations which perfectly match your input information.
The purple areas of the circle mark locations within a 20% error band (included because your input information may not be perfectly measured).
The grey area shows the parts of the world in daylight.
The white area shows where the sun is below the horizon.
affects the size of the matching output area. A small output area is usually desirable for a researcher, since it reduces the search space of a geolocation.
Note that the tool can be used to consider a range of inputs if the exact values are not known, since the outputs vary relatively smoothly with changes in input values.
Shadow Finder needs the date and time of the source image containing the shadow to be able to generate potential locations. This is not always available, particularly as social media sites usually remove this type of metadata (called EXIF data) before allowing an image to be posted. Researchers also need to be aware that the time and date an image is posted online is not often the same as when it was taken. HOWEVER Shadow Finder may still be useful as a "What If?" tool to exclude potential locations.
Using an object's height and the length of its shadow (or the angle to the sun) with the date and the time, the Shadow Finder code estimates the possible locations of that shadow. It plots all the possible locations (there will almost always be very many possibilities) on a world map as a bright line.
The length of the shadow fed into Shadow Finder can only be measured directly from the image if the camera is at right angles to the shadow. If not, calculations/adjustments will need to be made to correct for any other camera perspective.
Click the two tabs below to see the different views presented to the user as they find Shadow Finder on the GitHub website (Tab 1) and then run it in Colab (Tab 2):
The first tab above (GitHub Page tab) shows what the Shadow Finder page looks like on GitHub, before running in Colab.
The second tab above (Click on the COLAB INTERFACE tab) shows the Colab interface which will appear after the user Runs Shadow Finder in Colab using the Orange Button.
Date
Time
EITHER
Object and Shadow lengths
OR
Angle of sun elevation
The output map appears at the bottom of the page when Shadow Finder code has been run in Colab. Compare the Shadow Finder output for a given date and time with for the same point in time, shown below:
The two representations concur on the daylight distribution and sun position for the same time and date shown:
The centre of the world map is in daylight on both maps, with either side in darkness over the Pacific Ocean and (roughly) the Arctic Circle.
The sun is above the Atlantic Ocean to the west of Central Africa in both maps, with the locus of the Shadow Finder map centering on the location of the sun in the "" website map.
Data input is remarkably simple, as is running the tool if the user is familiar with GitHub. There is some theoretical understanding of the relationship between sun position, date and time required to use the tool effectively.
GitHub code such as Shadow Finder requires a GitHub account so that a user can log in to access and run the code.
Shadow Finder cannot find possible location points of a shadow from its length alone: the height of the object casting the shadow must also be known (or the elevation angle of the sun).
Shadow Finder does not provide the direction of the sun (azimuth angle) at each potential location for the given date, time and elevation angle of the sun. This would only be useful if it was possible to establish the orientation of the source imagery to allow confirmation.
Bellingcat is the provider of this tool, authored by , Bellingcat Tech Community Facilitator, Jordan Gillard, Thomas Ellmenreich and Boris Nezlobin.
Scroll down the Shadow Finder page to README section and click the bright orange button labelled "TRY IT ON COLAB". This launches ShadowFinder in an easy-access environment called Colab.
Scroll down to enter your shadow measurement details into the app. Input EITHER
a. both object height and shadow length (using the same units, whether thats mm, cm, m, feet, inches. Pixels in an image may be the easiest units - it doesn't matter because Shadow Finder just needs the ratio of object height divided by shadow length)
OR
b. the elevation angle of the sun in degrees
Definition: The elevation angle of the sun is the inverse tangent of the ratio of the height of an object for which the sun casts a shadow to the length of that shadow. Shadow Finder needs that angle in degrees (not radians, be careful with units here).
and then enter the date, time (and time zone you are using):
Click the dark Right Arrow button in the white circle in the left margin to Run the tool, as instructed by the sentence "Click to find possible locations that match the below information". [The white Left arrow on the square blue background is *not* a button you can press!]
If you get this warning message, click the "Run anyway" button if you're happy that the code isn't abusing your privacy.
Be aware that the code may not always run instantaneously and could 'queue' for a brief time. The elapsed run time will be shown whilst executing the code and the total time taken displayed upon completion next to the Run arrow button. The output takes the form of a world map (see top of page) showing all possible points satisfying the input conditions on date, time, object height and shadow length (the circular locus).
Change the input values at will and re-run as required.
Given a photograph containing an object and its shadow at a known date and time, find the possible locations at which this could be observable.
Given the angle of elevation of the sun at a known dat and time, find the possible locations at which this could be observable.
Elimination of a potential geolocation of an image dependent upon the shadows it contains, i.e. rule out a potential location because it doesn't lie on the Shadow Finder locus for feasible dates and times.
Data can only be input as numeric values into the text input boxes in Shadow Finder running in Colab.
Data Input Accuracy
Comparison of Outputs for the same time and date with different types of input Data. The input area on the right (for input as lengths) yields a narrower output area than the one on the left (for input as an angle). So for the same measurement error, two lengths give a more accurate output than one angle.
i.e. The elevation angle of the sun in degrees.
i.e. the height of the observed object and the length of the shadow it casts
Sophie Tedling.
