Extracting road lane boundaries using the perspective transform and its application to the obstacle detection on curved road

Abstract

ARV(Autonomous Road Vehicle) is very attractive subject in computer vision. Many researchers are studying for practical use of the ARV. The vehicle has to go ahead along the road path, detect an obstacle on its own path and control itself. Automatic cruise control is one of the promising field of the ARV. To go and stop automatically the vehicle has to know its own speed, the curvature of the road to control steering wheel, the safety distance which is related to its speed and object on its own lane to avoid collision. Active sensors like laser radar, millimeter wave radar and ultrasonic sensors can measure distance directly from the vehicle to the obstacle but can not recognize the road environment thus can not control steering wheel adaptively. In this paper, using the perspective transform between the real-world coordinate and image coordinate, estimated the road lane boundaries on curved road and detected the obstacle on its own path. This is helpful of automatic cruise control system and sensor fusion system.