Current News



Miniature glasses on praying mantises helps in study of robot vision

Amina Khan, Los Angeles Times on

Published in News & Features

The praying mantises, it turned out, did not suffer from the same problem.

"They were just totally unfazed by this manipulation and kept on striking when objects were near and not when they were far away," Read said. "So that told us something fairly profound. That was telling us praying mantises do not use correlation for their stereovision -- at least not the correlation of image contrast. They're doing something really different."

The scientists then took it a step further. They showed each eye a different, uncorrelated set of dots, but both with a moving target passing through them. Again, where the humans failed to perceive when the object was "close," the mantises succeeded.

"Clearly they were able to match up the moving target even though the dot pattern didn't match at all -- and that's something that humans can't do," Read said.

The scientists think it's because the insects are not trying to correlate all the details in the images, but instead are looking for changes in the light patterns over time. Basically, they're looking for only things that move; the rest of the picture's little touches don't really matter.

There could be a reason for these very different systems, Read pointed out. Human visual systems are very good at picking out the differences in still images, which might help them to better see through camouflage. On the other hand, based on their hunting style, praying mantises might be better served by a system that picks up only the movements of nearby prey.


The findings could help scientists design better computer algorithms for simpler visual processing systems in drones and other robots, the researchers said.

"We think the mantis 3-D algorithm, since it's simpler, could be implemented into processing systems that require less computing power," said lead author Vivek Nityananda, a behavioral ecologist at Newcastle University. "It could be implemented for example into lightweight robots."

(c)2018 Los Angeles Times

Visit the Los Angeles Times at

Distributed by Tribune Content Agency, LLC.



blog comments powered by Disqus