In this paper, we propose a system for observing rear and rear side areas of a vehicle by image processing. These areas are blind spots for drivers. To observe following vehicles on the driving lane and the passing lane, is very useful to avoid sudden collisions and accidents when changing lanes. Our system is divided into three parts, which are Lane Detection, Vehicle Detection and Vehicle Observation. In Lane Detection, a driving lane and a passing lane are detected by tracking lane-markers which appear from the bottom of images and approximating lane-markers from a polynomial approximation. In Vehicle Detection, the reliability is evaluated from the results of First Vehicle Search. If reliable areas are detected, these areas are tracked. This tracking algorithm also can be used even when following vehicles are changing lanes. In Vehicle Observation, the distance from a following vehicle to our vehicle is calculated, and then a relative speed is calculated from inter-frame movements. Finally, dangerous vehicles are detected from the relative speed and the position. If dangerous vehicles are detected, our system warns a driver about these dangerous vehicles.
Abstract