قالب وردپرس درنا توس
Home / Technology / CMU researchers see computer vision around the corner – TechCrunch

CMU researchers see computer vision around the corner – TechCrunch



Future autonomous vehicle and other machine intelligence systems may not need line of sight to capture unbelievably detailed image data: recent research by Carnegie Mellon University, the University of Toronto, and University College London has developed a "seeing around the corner" technique. "

The method uses special light sources, combined with sensors and image processing, to effectively derive or recreate extremely detailed images that are much more detailed than previously possible, without photographing or otherwise directly viewing them & # 39; 39; to have.

There are some limitations – so far researchers involved in the project have been able to use this technique effectively only for "relatively small areas," according to Professor Srinivasa Narasimhan of the CMU Robotics Institute.

This limitation could be mitigated by the use of this technique along with other techniques used in the field of non-line art-of-site (or NLOS) Computer Vision Research. Some of these techniques are already being used in the marketplace, including how Tesla's autopilot system (and other driver assistant technologies) uses reflected or reflected radar signals to see the cars immediately in front of the Tesla vehicle] The technique used in this new study is similar to that in a LiDAR system used in many autonomous vehicle systems (although Tesla is known to avoid the use of laser based image processing systems in its tech stack). The CMU and its partner institutions use ultra-fast laser light in their system, which is reflected off a wall to illuminate an object hidden around a corner.

Sensors then detect the reflected light as it bounces back, and researchers measure and calculate how long it took for the reflected light to return to the origin. Through a series of measurements and the use of information about the geometry of the target, the team was then able to reconstruct the objects with remarkable accuracy and detail. Their method has been so effective that it works even with half-agonizing materials, including heavy paper – another great advantage when it comes to using in environmental sensors that operate in real-world conditions.


Source link