An AI-enabled imaging system developed at MIT can readily analyze reflections to “see” around corners.
Self-driving cars have been in the testing phase for years now and many carmakers have semiautomatic driving modes, like Tesla’s AutoPilot.
Despite technological advances in recent years, manufacturers are still seeking to improve the safety of these vehicles, which are susceptible to confusion. Already, clever engineers have devised ways for algorithms to identify hazards on the road and better detect pedestrians.AI algorithm analyzes light reflections to 'see' around corners.Click To Tweet
Road safety is the major macro benefit of autonomous vehicles, as they are equipped with a number of sensors to detect obstacles in real time. Future self-driving vehicles could see around obstacles with edges and react much faster than a human driver could. What’s more, autonomous cars won’t fatigue or get emotional–things that cause human drivers to damage life and property.
Researchers at MIT have come up with a relatively simple and inexpensive imaging solution to see around corners, which works by recording and analyzing light reflections of objects on the floor.
AI and Penumbra to see Around Blind Spots
Researchers at MIT’s CSAIL (Computer Science and Artificial Intelligence Laboratory) reported they have developed a new imaging system, called Corner Cameras, to virtually do away with driving blind spots.
Thanks to an AI algorithm and standard RGB cameras, this system can analyze tiny details of light and shadows in order to deduce the position of objects, tracking them in a short period of time, determining their speed and trajectory.
Corner Cameras, which can also work with smartphone cameras, analyzes light reflections to detect people or objects that are hidden around corners and thus invisible to the naked eye.
“From standard RGB video recordings,” CSAIL team says in the abstract of their paper, “we use edge cameras to recover 1-D videos that reveal the number and trajectories of people moving in an occluded scene.”
Researchers will present the paper, titled Turning Corners into Cameras: Principles and Methods, at the International Conference on Computer Vision that will take place from October 22nd to the 29th in Venice (Italy).
CSAIL’s new system works in combination with various surfaces and with almost any lighting. The exception, of course, is complete darkness, which could be a problem on some roads. Another issue is frequent changes in lighting, for example in the case of smaller clouds covering the sun.
Corner Cameras could be used in conventional cars as a driving-assistance system, and also suggests a future where autonomous vehicles will be able to react much more quickly to immediate danger than a human driver. The technology could also be adapted to other types of simpler vehicles such as wheelchairs.
Will Corner Cameras make you feel better about getting in the passenger seat of an automous vehicle? If not, what would?