New system allows vehicles to 'see' objects hidden behind corners

Such systems are intended in the future to be used in autonomous vehicles without a driver, but also in other uses

Engineers in the US have developed a new imaging system that - with the help of laser beams - produces images of invisible objects, which are hidden behind corners.

Such systems are intended for future use in autonomous driverless vehicles, but also in other uses, such as rescue teams, to see people covered in debris in the event of a natural disaster.

The researchers, led by Assistant Professor Gordon Weetstein of the Department of Electrical Engineering at Stanford University in California, published the paper in the journal Nature.

"It sounds like magic, but the idea of ​​depicting things that are out of line of sight is really possible," Weetstein said.

In this way, if e.g. an autonomous car moves through the narrow streets of a city and prepares to make a closed turn, but behind the corner a (invisible) child plays with his ball, the "smart" vehicle will see it back and will stop in time.

These researchers are not the only ones in the world who are developing hypersensitive laser beam technologies that "twist" around corners and reflect on hidden objects to betray their presence. But the novelty of the new method is that it incorporates a new highly efficient computational algorithm, which processes laser data in order to isolate "noise" and produce images of hidden objects.

 

The laser is located next to a hypersensitive photon detector, which can detect even a single particle of light. When the pulses of laser light are sent to the target (eg the corner wall) and reflected back, the detector "catches" the photons that have fallen on the hidden objects and then the algorithm undertakes to "assemble" the light particles in a final image of the invisible thing.

At the moment, the first stage - that is, the collection of laser photons that have fallen on the hidden object - takes from two minutes to an hour, depending on the lighting conditions and the reflectivity of the hidden object. The second stage - the processing of this data by the algorithm - is done almost in a flash, in less than a second.

The goal of the researchers is to improve the technique, so that it creates images almost immediately, which is absolutely necessary if it is to be applied in real driving conditions by autonomous vehicles, where the reactions of the car must be immediate.

Today's autonomous cars use LIDAR systems to navigate. The new algorithm could be integrated into these systems in the future. But in the meantime the new system needs to be improved to work better in daylight and with hidden objects on the go, like a child running around the corner.

As the tests of the new system showed, if placed in an autonomous vehicle today in its current stage of development, the driverless car would easily detect things like road signs behind the corner, but would have difficulty with people if they were not wearing clothes. with reflectivity.

  

Source: News247