Researchers create technology that can ‘see’ around a corner via camera images
Computer scientists at the Massachusetts Institute of Technology have developed a technology that can map moving objects or people hidden behind a wall around the corner by analyzing light and shadow information on camera images.
The technology works by analyzing the light information on the floor near a corner, such as a corridor with a T-junction. When images are taken of the floor near the point where you have to turn the corner, a computer program can detect changes in the light on the floor. Everything reflects light, and photographing the fluctuations in light intensity on the floor can give you an idea of what’s around the corner.
These changes in luminosity are imperceptible to the human eye, because the fluctuations occur in only 0.1 percent of the light that is reflected. By enhancing the colors of the images with a computer, certain colored pixels can be linked to certain objects or subjects around the corner. Someone with a green shirt will reflect green light, or the existing light source will be blocked so that a dark pattern is created. The movements of people who are around a corner can be mapped with this. It is in fact about differences in the pixels, so that it also works in real-time, one of the researchers reported to Wired.
The technology is relatively simple, also works outside and in the rain, and works with images from, for example, a Sony A7s system camera or an iPhone 5s. However, there are still some limitations. For example, the camera still has to be at a fixed point, although the researchers are now also trying to get the system working with moving images. This could make the technology usable for use in self-driving cars. The researchers believe that this could eventually work if the car is still quite far from a corner. A laptop is still needed to analyze and process the image information, but the scientists believe that the system should also be able to work completely on a smartphone, although that has not yet been done.
The research is published in a paper titled Turning Corners into Cameras: Principles and Methods.