Imaging and sensing are key to autonomous vehicle technology and an area of focus has been with 3D laser scanning systems like LiDAR (light detection and ranging), which self-driving vehicles deploy for obstacle detection and avoidance to navigate safely through environments, using rotating laser beams.
However, current imaging technologies can be improved to enhance the safety features. This is the subject of research conducted at Boston University. Here technologists have developed a new method to allow the artificial intelligence component in autonomous vehicles to see around corners.
To achieve this a combination of scanning lasers and highly sensitive cameras has been tried, with some success. However, what the Boston researchers have managed is to achieve the ability for an autonomous vehicle to sense what is around a corner without the need for advanced optics. Their technology is based on a standard digital camera plus a specially developed algorithm described as a “computational periscope”.
The technology works in a way similar to toy periscopes: a device composed of twin mirrors or prisms enables a person to see objects that are blocked by obstacles.
The new technology does not rely on mirrors, but the principles are similar. The Boston algorithm uses the fact that light bounces off wall-like structures with varying patterns, being able to assess the degree of disordering that different objects create. The artificial intelligence has been created to “unscramble” the scattered reflections.
Tests are underway to see how well this works in practice, with the expectation that the algorithm can be used to monitor hazardous environments and navigation, perhaps aiding autonomous vehicles to see-around-corners.
The technology has been presented in the journal Nature. The research paper is titled “Computational periscopy with an ordinary digital camera.”
