Once, super heroes were able to see into closed rooms. Researchers at Stanford Computational Imaging Lab developed a non-line-of sight imaging technique that allows you to see the physical objects inside a room with just one point of laser light.
Advertisement
Imaging using non-line-of sight (NLOS) is not a new concept. This clever technique has been perfected over time in research laboratories to produce cameras that can see around corners and take images of objects that are not in the camera's field of view or blocked by obstacles. The technique was previously used on flat surfaces such as floors and walls, which are within the line of sight of the camera and obstruction. The camera emits light pulses that bounce off of these surfaces. These light pulses are usually generated by lasers. Algorithms use this information to calculate the time it took for these reflections to return in order to create an image of what the camera cannot see. Although the results may not be high-resolution, they are usually sufficient to identify the object.
This is a clever technique that could one day be used to detect potential hazards in corners before they are visible to the passengers of autonomous cars. It would improve safety and avoid any dangers. The current NLOS techniques are limited by the need for a reflective surface that can reflect light from hidden objects. It is almost impossible to see what is inside a room from the outside, or at least it was not until recently.
Stanford University's Computational Imaging Lab developed the keyhole imaging technique. It is named for the tiny hole that allows you to shine a laser beam through a small hole, which creates a single light spot on the wall. Similar to previous experiments, the laser beam bounces off walls and objects in the room before reflecting back to the camera. The single-photon avalanche detector measures the time of their return.
The new keyhole imaging technique can't see what an object in a room is seeing if it is stationary. Researchers discovered that a moving object coupled with pulses from light from a laser can generate enough data over a long exposure time to allow an algorithm to create images of what it is seeing. Although the quality of the images is not as good as previous NLOS techniques it still gives enough detail to allow for an educated guess about the object's size and shape. Although a wooden mannequin may look like a ghostly angel when paired with an image recognition AI that is properly trained, it makes it very possible to determine if there was a human or human-shaped object in the room.
G/O Media could get a commission Editor’s choice Anker Wireless Charging station Charge your phone and keep an eye on it simultaneously
High compatibility with smartphones and watches
Case friendly
Buy Now
This research could eventually allow police and military to evaluate the risk of entering a room without actually breaking down the doors or storming in. This new technique could be used by autonomous navigation systems to detect hidden dangers before they become a problem in environments where NLOS techniques aren't feasible.
