Scientists from the United States have presented a new technique: it, with the help of a laser, which is directed into a peephole or keyhole, allows you to visualize an entire room. The quality of the final image is improved by the AI model.
Researchers at Stanford Lab have perfected non-line-of-sight imaging techniques. Now scientists only need one point of laser light to hit a room. It can be used to see what objects are inside.
The core of the method is a technology that has been used for many years to create cameras that can see around corners and generate images of objects that do not enter the camera’s field of view or are blocked by obstacles. Previously, this technique used flat surfaces such as floors or walls that were in the line of sight of both the camera and an obstructing object.
A series of light pulses emanating from a camera, usually lasers, bounce off these surfaces, then bounce off a hidden object, and then return to the camera’s sensors. The algorithms then use information about how long it took for the reflection to create an image of what the camera cannot see. The results are usually of poor quality, but even this is enough to identify the object.
Keyhole imaging is so named because it allows you to see objects inside a closed room through a tiny hole (such as a keyhole or peephole). A laser beam enters the room through it, creating a single point of light on the wall inside the room. Then the light bounces off the wall, off the object in the room, off the wall again. Countless photons are reflected back into the camera, which uses a single-photon avalanche photodetector to measure their return time.
When an object hidden in a room is stationary, the new keyhole imaging method simply cannot figure out what it is seeing. But the researchers found that a moving object paired with pulses of light from a laser generates enough useful data over long exposure times for an algorithm to create an image of it.
Researchers have improved the quality of AI recognition. It can even detect blurry images of a person or a closet by adding a picture from an image database that already has similar models.