One of the unexpected new uses of artificial intelligence is to help map numerous icebergs. Faster and more accurate than the human eye, this AI will now make it possible to map and track the evolution of the largest moving icebergs.
Traditionally, icebergs have been mapped by humans using satellite images, but this method is neither very accurate nor very fast.
That's why scientists from the UK's University of Leeds, working in collaboration with the European Space Agency (ESA), have developed an algorithm capable of mapping icebergs accurately and very quickly from satellite images. In fact, it takes AI just one hundredth of a second to do a job that would take a human several minutes.
A computer has been trained to identify the precise contours of icebergs in images sent by the Sentinel-1 satellites. The algorithm, dubbed U-net, proved more effective than any previously programmed.
So far, the system has been tested on images of seven different icebergs, ranging in size from 54 to 1052 sq km. The first tests were carried out in the North Atlantic. Others should follow in different regions.
Iceberg tracking is crucial for observing and analysing the polar environment, as well as the ice melting process associated with rising sea levels. Ultimately, this AI could help to improve maritime safety, preventing collisions between ships and icebergs, but also to protect coastlines, to assess rising sea levels, and quite simply, to aid scientific research. – AFP Relaxnews