Communications of the ACM
Car Safety System Could Anticipate Driver's Mistakes
By monitoring a driver's head movements and watching the road ahead, a computer can guess the driver's intentions. Here, the computer predicts the highest probability for a left turn at the next intersection.
Credit: Robot Learning Lab
Cornell University researchers have developed an algorithm-based automotive system that analyzes a motorist's body language in context to what is happening outside the car to predict if the driver will turn, change lanes, or continue going straight.
The system merges driver anticipation data with radar or cameras to locate other vehicles in order to warn the driver when an expected action could be dangerous. The system also uses global-positioning system information and street maps to issue an illegal turn message.
Cornell professor Ashutosh Saxena and colleagues developed the system by recording videos of 10 drivers and videos of the road ahead for 1,180 miles of city and highway driving over two months. A computer with face detection and tracking software analyzed head movements and learned to associate them with turns and lane changes to enable the final system to predict actions the driver may take, achieving 77.4-percent accuracy at an average of 3.53 seconds in advance.
The researchers say the system still needs refinement, because 6 percent of the time its face-tracking ability was confused by shadows of trees and other lighting variations or by drivers interacting with passengers. If drivers rely on short-term memory of traffic conditions and do not turn their heads to check, the system might rely solely on tracking eye movements, the researchers note.
From Cornell Chronicle
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found