A stereovision based rear-end collision warning system on mobile devices

In this paper, we tackle the challenge of developing a stereovision-based driving assistance system for collision avoidance on a mobile device. This novel system is aimed to work in an urban environment, where a driver experiences lots of stop and go situations. In cities, rear-end accidents are the most common type of accidents, since a momentary lapse of concentration can lead to unsafe headway. The system detects and tracks the vehicle in front of the host vehicle and issues a warning if a crash is imminent, such that the driver has sufficient time to brake or take another action to avoid the accident. Complex functions were developed for lane detection and tracking and vehicle detection and tracking. The rear-end collision warning is based on the computation of Time-To-Collision and Time-Gap taking into account host vehicle speed, relative speed and relative acceleration. Stereovision-based algorithms for driving assistance are computational intensive and mobile devices as they are today have reduced resources. Our algorithms run in real-time, at 20-22 frames/second and at the same time they are robust and have high accuracy.