Measuring Distance with Less Light
There are many situations where you want to measure the distance to an object, such as surveying or controlling an autonomous car. A common tool for those situations is the lidar rangefinder, which reflects laser light off of objects to make the measurement. Researchers at MIT have recently made some clever advances to the system, enabling it to use significantly less light, which should have some interesting benefits.
A typical lidar system will repeatedly fire laser pulses at a position, until it gathers enough consistent data to be confident in the distance, and moves on to another position. The new MIT system however only accepts one photon before moving on, but records number of pulses it fired before receiving the photon. This allows it to generate a map based just on that data, roughly indicating the reflectivity of different objects. Of course, capturing only a single photon puts the system at risk of being fooled by a stray photon, from another source. To correct for that the researchers are applying a statistical trick that takes advantage of the fact that such photons follow a pattern known as Poisson noise. Instead of just filtering out the noise pixel-by-pixel, this system considers how much filtering was required in adjacent pixels, as they will likely have similar reflective properties, at the same depth.
Altogether, this new system should be able to generate a depth map with just a hundredth the number of photons a conventional lidar system uses, and generate an image with one nine-hundredth the number of photons. This should result in energy and time savings, and should prove useful in low-light situations.