In the not-too-distant future a driver will zip down the street, coffee cup in one hand, four squirming kids in the back. She'll take her eyes off the road as she tries to settle the kids down, right as she approaches a busy intersection.
Before she zooms headlong into traffic, however, a special camera in her car will notice the situation. A tell-tale sound will let her know what's about to happen, so she can get her attention back to where it needs to be - on the road.
Such a device isn't the realm of science fiction. Scientists at Los Alamos National Laboratory have developed software that could make the technology standard issue - for about $200 a pop - on new cars in the next five to 10 years, said Lakshman Prasad, who is working on the project.
"We thought it could be a tremendous help if computers could see around you and tell you, 'Hey, you're about to jump a stop sign,' " Prasad said. "To do that we had to learn how to mimic how humans so cleverly process visual information and know what to pay attention to."
Prasad and Alexei Skourikhine, another lab scientist, have been developing the technology since the late 1990s and say their work is just about ready for prime time.
They're looking for commercial partners to take it out of the lab and into a busy driver's car, they said.
"There's information overload today," Prasad said. "We have lots of devices around us like cameras and computers to capture data, but nothing to process it all. This technology can help people sort out information when they're tired or overloaded. It has a lot of promise."
The device made of the software and special cameras could also help high-speed trains avoid hitting cars by watching tracks far ahead and identifying obstructions, the two said.
"The technology is much more sensitive than the human eye - especially after several hours of driving," Prasad said.
Transmitting images from fixed cameras raises even more possibilities, Prasad said.
The software converts complex visual data into a series of simple shapes that can be easily transmitted over cell phones or the Web. It also recognizes what the shapes are based on patterns programmed into the system by Prasad and Skourikhine, they said. "A truck driver might be interested in a situation coming 50 miles down the road," Skourikhine said. "Because these (simple shapes) don't use a lot of bandwidth, cameras could transmit that quickly to the driver's truck so he could see what he's approaching."