IN A LOCKED room in a busy city, some terrorists are holding a hostage. The curtains are mostly drawn, cutting off any direct line of sight for those outside. In a building across the street, a team of engineers are set a task: they can have whatever equipment they need, but they must paint as clear a picture as possible of what is happening inside the room.
This was the challenge given in 2015 to Daniele Faccio, then at Heriot-Watt University, in Edinburgh, by Dstl, a British government defence laboratory. He and his team eventually found a way to see around corners from a distance of 50 metres—which was reckoned impressive at the time, even though the system they devised could detect only the motion and position of hidden objects, rather than taking pictures of them. Now, however, Xu Feihu and Pan Jianwei of the University of Science and Technology of China, in Hefei, have blown that record out of the water. As they describe in the Proceedings of the National Academy of Sciences, they have managed to look around corners from a distance of well over a kilometre.
Non-line-of-sight imaging of this sort relies on two principles. One is that objects are visible to an observer if light bouncing off them makes its way to that observer’s eyes or instruments. The other is that at least some light reflects off all but the blackest, most absorbing surfaces. The upshot is that something hidden from an observer’s line of sight might nevertheless be visible if it is sufficiently near a wall which can serve as a reflecting surface. In this case, the observer can illuminate the wall with a tightly focused beam of light (in practice, probably a laser), knowing that some of the beam’s light will bounce off the wall to illuminate the concealed object, and that some of this illumination will, in turn, be reflected back whence it came via the wall. The fraction of the original beam returned by this trilogy of reflections may be minuscule, and the information it contains may appear hopelessly jumbled. But sufficiently smart mathematics can turn it into an image of the thing it bounced off.
Mirror, mirror, you’re the wall
Dr Xu and Dr Pan conducted their trial at night, to minimise the amount of background light that might have interfered with the results. Their targets, a dummy of a human being on one experimental run and a giant “H” on another, were concealed behind a barrier in an apartment in a block of flats in Shanghai. Their laser and receiving apparatus were in a second apartment block 1.43km away. The receiving apparatus, an instrument called a single-photon avalanche diode (SPAD), was, as its name suggests, so sensitive that it could detect and count individual photons, the particles of which light beams are composed. This was just as well, for, of every seven million billion photons fired across the gap by the laser, only a single one returned.
Because each firing of the laser yielded so little information, the researchers had to take many shots to build up an image. To that end, they imagined a grid on the target wall, 64 dots wide and 64 deep. They fired the laser at each dot in turn, and then fed the data from the SPAD into an algorithm capable of reconstructing, albeit fuzzily, an image of the hidden object (see picture).
The military applications of this technology suggest themselves. They were, after all, why Dstl sponsored Dr Faccio in the first place. But others are also interested. America’s space agency, NASA, has paid for such work in the past in the hope of putting a laser on a satellite orbiting a distant world. This would permit the photographing of the otherwise-invisible interiors of caverns on the surfaces of moons and planets. And, in a more practical vein, engineers in the autonomous-vehicle industry would be keen on a technology that let their cars spot other motorists blithely speeding around blind corners.
For now, such applications remain far in the future. Capturing the experimental data in Shanghai took several hours, which is of little use either on the road or in fast-moving situations like hostage-taking. The amount of light lost between bounces also puts a limit on how far away an object can be from the reflecting wall before the technique stops being useful. The dummy used by Dr Xu and Dr Pan was 75cm from the wall, which is probably near that limit. These caveats aside, however, performing the trick over a distance of almost 1½km is a staggering advance on previous efforts. It would not be surprising, says Dr Faccio, now that it is known what is possible, if that record, too, were broken.■
This article appeared in the Science & technology section of the print edition under the headline “Round the bend”