Sticking a small patch on a large object like a plane can hide it from artificial intelligence systems trained to spot objects in drone footage. The technology could help conceal military assets from drone surveillance, say Ajaya Adhikari and Richard den Hollander at the Netherlands Organisation for Applied Scientific Research.
They and their colleagues used an聽AI that generates a pattern to聽confuse a drone surveillance system called the YOLO object detector, which spots military objects in aerial images.
The researchers overlaid several patterns of different sizes on aerial photographs, and found that a certain pattern that looks a bit like聽colourful tie-dye prevented the聽object detector from spotting jet聽fighters in these images.
Advertisement
The most effective one was about 10 times smaller than the plane, and it worked best when placed atop the aircraft in the image rather than beside it. This size patch reduced the accuracy of the YOLO detector from 94 to 38 per cent.
Sticking a patch like this onto an聽object in real life could conceal it聽from surveillance, say the researchers. 鈥淲e believe this technology will be applicable in other security domains where adversaries have an interest in concealing objects in images,鈥� say聽Adhikari and Hollander. For聽instance, similar technology developed last year conceals faces from surveillance cameras.
The researchers are also studying defences against such patches. 鈥淭he聽detection of objects on the ground and the use of camouflage to prevent detection are in a continuous competition,鈥� they say.
Reference:
