Building safer roads

Close calls on the road can do more than just smarten up bad drivers.

UBC road safety expert Tarek Sayed has devised a way to automatically track and analyze “near misses,” using the findings to improve highway operations and traffic design, and ultimately reduce the number and severity of car crashes.

“This is a new technique that we first developed at UBC and is currently being applied in several projects worldwide,” says Sayed, professor of civil engineering in the Faculty of Applied Science.

Sayed is currently running pilot projects of the “computer vision and automated safety analysis” in Vancouver, Edmonton, Penticton, Calgary, Oakland, Cairo and Kuwait City.

The upfront investment for the video monitoring is low, since it only requires an ordinary $300 digital camera mounted on a pole to track motorists. The video camera is then left to record between one and three days, depending on traffic volume.

Sayed then analyzes the video data using a software program he developed with UBC graduate students and research associates. The program’s algorithms isolate the types of potential collisions and their degrees of severity.

“As well, we use traffic conflict techniques such as an extrapolated ‘time-to-collision measure’ to better understand the split-second responses and behaviour of drivers during these conflicts.”

Partnering cities have embraced this innovation, says Sayed. “The biggest advantage of using video recording and automated analysis is that it reduces the amount of required data from years to mere days.”

Traditionally, road improvement programs evaluated the types and rates of collisions over a period of two to three years, using records from police reports and insurance claims. In some cases, accident statistics would be supplemented by human observation—people hired and trained to record and evaluate the frequency and type of traffic conflicts.

“Historical collision data and human observation are not always reliable or even available,” says Sayed. “In contrast, video sensors provide rich, detailed, inexpensive and permanent observations of traffic scenes.”

He adds that decision makers can then more easily set priorities for highway improvements or design changes when working with robust and precise data. “It’s possible to assign an average cost per incident and evaluate the safety benefits.”

The city of Edmonton, for example, tackled one of its top crash-prone locations after seeing Sayed’s data.  A merge ramp on the Yellowhead/Victoria Trail intersection was particularly dangerous, directing drivers into high-volume traffic. Now the location has a right-turn traffic light, reducing the number of collisions along that stretch of highway by 95 per cent.

And in Oakland, Sayed evaluated motion patterns for “pedestrian scramble,” a mode that halts traffic in all four directions. At the “Walk” sign, pedestrians on all four corners of the intersection can cross the street, even diagonally.

“There were significantly higher potential conflicts before the pedestrian scramble.”