10 million photos train AI model, scientists find a new method of underwater location

CTOnews.com, July 11 (Xinhua)-- Satellite systems such as Beidou and GPS can accurately locate the surface, but there is nothing they can do underwater. Scientists at the University of Illinois at Urbana-Champaign recently found a way to navigate underwater without a positioning system by analyzing the polarization patterns of underwater light.

CTOnews.com Note: the positioning system mainly depends on the signals transmitted by multiple satellites to determine the location of users on land, but this kind of radio signal is very weak and unstable underwater, so it can not be used in submarines or underwater exploration activities.

The university team used underwater cameras with special optical elements to take about 10 million photos at multiple locations, under different conditions, on different dates, at different depths and at different times of the day.

The team developed an artificial intelligence algorithm and used the above photos for training, which can locate underwater at a depth of more than 300 meters, with a recognition accuracy of 40-50 kilometers.

CTOnews.com attaches the address of the paper here, which can be read in depth by interested users.