Scientists from California have developed a method to make hidden things that are right around the corner visible for cameras. To do this, they scan surfaces in the field of view with a laser and analyze the light reflected by hidden objects. According to Gordon Wetzstein and his colleagues from Stanford University, the technique could be used in medicine, in the perception of robots and in autonomous driving.
Day and night, scientists are making breakthroughs in different fields. My personal favorite is the invention of playing live casino games with augmented reality or virtual reality devices. It is like being in Las Vegas, but from your own home.
Anyway, this new camera may sound like magic, but the idea of imaging hidden things is actually feasible according to Wetzstein. The biggest challenges are the extremely weak signal of the scattered light and the tremendous computing power that was previously necessary for the reconstruction of images.
The researchers solved the first problem with a very sensitive detector that can detect single photons (light particles). Thanks to a combination of efficient algorithms, they were also able to reduce their computing requirements greatly. The image calculation is possible even on a standard laptop.
At the center of the processing is a so-called Wiener filter, which performs a noise suppression. To do a measurement, a laser pulse is directed to a point and the reflected light is recorded at the same point. From this, finally, a hidden object can be reconstructed three-dimensionally.
In investigating how powerful their system is, the group gathered around whetstone over one billion voxels (three-dimensional pixels). That was done in a thousandth of a second. The scientists are therefore confident that their method can be applied in real time.
The researchers conclude that their system enables imaging of hidden things “with conventional hardware at much higher speeds, with smaller memory requirements and lower power consumption, over a longer range, with ambient lighting and with higher resolution than any existing approach we know about.
They are already working on improving their process. So a stronger laser could be used to reduce the recording time of the scanned object. This would have to radiate in the short-wave infrared range, so as not to endanger a person’s eyesight.
To reduce the computation time, the multiple parallel processing algorithm could also be executed on a graphics processor. Carlos Fernández López from the Karlsruhe Institute of Technology (KIT) considers the topic very important, for example to make autonomous driving safer than it is at the moment. However, so far the method has only been tested on objects that were a few centimeters apart.
In the event of a dangerous situation when applied to autonomous driving, the vehicle should detect and stop at least 20 meters before it reaches the obstacle, depending on the driving speed. Nevertheless, the study is an important step forward in the reconstruction of objects outside the field of vision.