The children are fighting in the back seat. Someone just dropped their juice on the floor. The driver tries to stay focussed on the road. But with the commotion, the car drifts into the next lane. An angry horn blast alerts the driver to an oncoming car. An accident is avoided but it's too close for comfort for most drivers.
Robert Laganière, an engineering professor at the University of Ottawa, is developing better ways to warn drivers of danger to help avoid such close calls. His research makes use of cameras and advanced artificial intelligence to keep extra eyes on the road.
"Safety features have proven to be a highly marketable feature for automakers," says Dr. Laganière. "The major technological challenge is developing new and better machine learning algorithms that can process visual information in real time—at the speed of a moving car—to recognize objects and provide accurate warnings."
These algorithms—a set of rules that a computer refers to when processing new data—are loaded onto a microprocessing chip that is installed on the car and connected to on-board warning systems. Dr. Laganière has been working with a private company to create marketable applications for the research done by his team at the VIVA Research Lab, which focusses on video processing and computer vision.
Using an NSERC Engage Grant that fosters such collaboration, Dr. Laganière has teamed up with CogniVue Corporation—a maker of driver awareness technology. The expertise of Dr. Laganière and his team has helped the company improve their software. Lane-detection technology already exists in some on-board cameras and Dr. Laganière is working to make them more reliable, especially in challenging weather conditions like a Canadian winter.
"Through this collaborative effort with Dr. Laganière and his team, CogniVue Corporation has amassed extensive knowledge in computer vision algorithms for driver safety systems", says Luc Martel, Director of New Technology Development at CogniVue Corporation. "Having these leading researchers work hand-in-hand with our embedded software team has proven to be instrumental in accelerating the development of these automotive vision systems."
Because of the potential for these technologies, the partners plan to continue working together even after the grant expires. Dr. Laganière sees the potential for a number of applications. He points to cameras that could help detect when you are following a vehicle too closely. Cameras could be used to read and display road signs. Cameras in the back of a vehicle, which many newer models already have, could send a warning when you are about to hit an object.
"It's even possible to have cameras aimed at the driver and then develop algorithms that recognize signs of fatigue," says Dr. Laganière. "The collaboration has been a good opportunity to get to know each other under the auspices of a first research project and there's tremendous potential to move more research from the laboratory to the marketplace."