If only drivers could see through walls, blind corners and other dangerous road junctions would be much safer. Now an augmented reality system has been built that could just make that come true (see video, above).
The prototype uses two cameras: one that captures the driver鈥檚 view and a second that sees the scene behind a view-blocking wall. A computer takes the feed from the second camera and layers it on top of the images from the first so that the wall appears to be transparent.
This makes it simple to glance 鈥渢hrough鈥� a wall to see what鈥檚 going on behind it. But the techniques needed to combine them were challenging to develop, says of Carnegie Mellon University in Pittsburgh, Pennsylvania.
Advertisement
Altered images
The view of the hidden scene needs to be skewed so that it looks as if it were being viewed from the position of the person using the system. The system does this by spotting landmarks seen by both cameras: the one seeing the hidden view and the one with the same view as the user.
Sheikh and his colleagues also had to develop software that transforms moving objects in the images to avoid distortion.
Ultimately, the team want to build the system into a car. An onboard video processor would tune into a wireless feed from a roadside camera with a view of the hidden scene, such as a stretch of road behind a blind corner, and project the image of the hidden scene onto the windscreen rather than a monitor.
The project is funded by , a car parts manufacturer based in Kariya, Japan.
Future view
鈥淚t鈥檚 an interesting peek into the future,鈥� says of the University of South Australia in Adelaide. He points out that many cities already have networks of CCTV cameras that could provide footage of hidden scenes.
Such a network could be supplemented by images from cameras mounted on many cars, says Shiekh. The Carnegie team is working on software that integrates feeds in footage from such sources into the system.
But Thomas adds that several formidable hurdles will have to be cleared before the technology can be used on public highways. Fast, powerful data processing and communication would be required to make the system work usefully in a moving car in real time.
鈥淚鈥檓 not saying it鈥檚 impossible,鈥� says Thomas. 鈥淏ut I don鈥檛 see it in a car anytime soon.鈥�
The prototype system was presented at the in Orlando, Florida, this week.
Jim Giles tweets at
