Satellite imagery can’t usually map urban areas in the rain due to its inability to image though clouds, or at night since it is dependent on the sun for illumination. Radar satellites on the other hand can image in all weather and even in total darkness, however the images from them are difficult to interpret and need specially trained analysts to make sense of. By teaching an artificial intelligence algorithm the rules of physics, the researchers were able to extract human like performance in map generation.
The researchers simulated radar scattering from different urban areas, and trained the algorithm to develop an understanding of it. This way, even if the scene was new to the algorithm, by developing a human like intuition and deduction of radar physics, it is able to correctly map the area. This all weather mapping capabilities are crucial in case of natural disasters like hurricanes, where the ability to quickly track damage is crucial in saving lives. Or in military applications, where the ability to maintain surveillance of bunkers and encampments being built at night by insurgents is mission critical.
We are trying to get this development talked about in media. If you would like to draft an article featuring our work please contact me or reply here!
I have taken the liberty of drafting a sample article, with infographics and images. Please find it here:
https://drive.google.com/open?id=1WAg2v6zKaqJWjIa6wj4r_hWNYWsRfqJqbCiTZZ05AXQ