Artificial eye for car co-pilot

Tuesday, May 4, 2010 @ 05:05 PM gHale


All it takes is a half second.
In that time, a driver can swerve to avoid a fatal accident, a deer, or even slam on the brakes to avoid a child running after a ball. The catch is, though, the driver must perceive the danger.
That is where a rapid alert system comes into play. This type of system can help mitigate risks, fatalities and severe injuries from road accidents, said Professor Shai Avidan of Tel Aviv University’s Faculty of Engineering. He is currently collaborating with researchers from General Motors Research Israel to keep cars on the road and people out of hospitals.


Advanced algorithms can help engineers develop a system where cameras mounted on cars can detect threats, and alerting drivers to make split-second decisions.
The challenge is to develop a system that can recognize people, distinguishing them from other moving objects — and to create a model that can react almost instantaneously, said Avidan. Ultimately, he is hoping computer vision research will make cars smarter, and roads a lot safer.
Cars all have engines, seats, and steering wheels. But autos are becoming much smarter with products such as the smart camera system by MobilEye, an Israeli startup company. Avidan was part of the MobilEye technical team that developed a system to detect vehicles and track them in real-time.
He is now extending that research to develop the next generation of smart cameras, which are aware of their surroundings. His goal is a camera capable of distinguishing pedestrians from other moving objects that can then warn the driver of an impending accident.
The challenge is in the development of a method that can detect and categorize moving objects reliably and quickly. By combining powerful algorithms to recognize and track objects, this tool could double check for vehicles in the driver’s blind spot, help swerve when a child runs into the street, or automatically block your door from opening if a cyclist is racing toward you, Avidan said.
Eventually, he hopes cameras will be able to recognize just about anything moving through the physical world. The underlying technology could also help in surveillance to detect a potential intruder.
Previously, detection systems used radar, which is expensive and not particularly sensitive to human beings. A smart camera fuelled by a powerful chip, on the other hand, could detect the activities of people and animals, and prompt the car to react accordingly, braking more or even locking the doors.
So far the technology works on infrared, greyscale, and color cameras. “Cameras are quite dumb machines unless you know how to extract information from them,” Avidan said. “Now, as the price of cameras drop and computer power grows, we’ll see more exciting applications that will keep us safe and make our lives more comfortable.”



Leave a Reply

You must be logged in to post a comment.