Will drones soon be able to see as well as birds do? Achintya K. Bhowmik, vice president and general manager of the Perceptual Computing Group at Intel Corporation, recently published a paper describing progress on that front. His paper on “sensification of computing” outlines research to enable machines to “see” their environment in three dimensions and in color, then use the differences they perceive to interact with their environment—just as people and animals do.
Drones like DJI’s Mavic Pro already can identify and track a person and respond to basic hand gestures. Researchers at Intel, which recently bought Movidius, the company that developed the technology at the core of the Mavic Pro, are working to turn two-dimensional image recognition systems into three-dimensional ones. That requires a second camera mounted at a slight offset from the first. A processor compares differences in shape and color to determine an object’s angle relative to each camera, which allows the system to triangulate how far away the object is. It’s the same way animal vision works, except with artificial cameras and processors instead of eyes and brains.
Intel’s RealSense technology is part of a drive to create machines with “intelligent capabilities,” says Bhowmik via email, meaning ones that can make their own decisions based on sensory data. He says the technology should also be useful for virtual and augmented reality devices. He predicts that artificial systems will eventually outperform an eagle, an animal with famously good vision, when it comes to depth perception.
“While we are still grasping the amazing capabilities of the human sensory system, including the transduction organs and the cerebral cortex, in several areas we have built sensors and artificial intelligence technologies that have already surpassed human capabilities,” he said. In addition to making better visible-light and infrared scanners, accelerometers, and gyros—refinements of what’s already on the market—he says Intel is exploring other technologies that should make machine vision better, although he declined to say what they were.