Detecting the presence of an object outside the line of sight and anticipating its movement are critical factors for many systems, starting with autonomous cars. A traditional camera suffers the same limitations as our eyes, but the CornerCameras platform developed by CSAIL at MIT can obtain information about what is hidden around a corner, using light reflections on the ground as a reference.
Let's imagine a camera capable of seeing around a corner. Think about its potential in dangerous areas, emergency situations, or environments that could become everyday in the future, such as streets full of self-driving cars. The idea is not exactly new. In 2012, MIT developed a camera equipped with laser technology that could reconstruct an object by firing pulses and using walls as reflective surfaces. The latest news brings us back to the prestigious institute, this time thanks to the efforts of CSAIL Computer Science and Artificial Intelligence Laboratory, but its brand-new creation does not rely on laser beams, but on something much more subtle.
How CornerCameras Works
Its official name is CornerCameras, and it is a system that combines camera recording with a robust algorithm. Suppose you are approaching a corner that blocks a couple of objects you cannot see. Those objects reflect a very small amount of light on the ground within your line of sight, creating a faint penumbra. By obtaining a video of that penumbra, CornerCameras creates a composition of one-dimensional images that reveal data about the hidden objects. If they are in motion, the system helps determine details including position, relative speed, and the moment when one object passes in front of another.
Challenges and Next Steps
It goes without saying that CornerCameras faces dozens of challenges. Complete absence of light, sudden changes in illumination, and even rain present difficulties. Upcoming studies will seek to develop specific solutions for each case, while the first evaluations will be done with a wheelchair. If the result is positive, their plan is to adapt the system to other vehicles.
