Road traffic incidents analysis has shown that 52 % of them are caused by a collision between two vehicles or between a vehicle and an obstacle. As a result, in the context of Intelligent Transportation Systems, onboard road obstacles detection is an essential task in order to reduce the gravity of accidents: on the basis of an obstacles detection device, one could design driving safety assistance systems such as emergency braking, collision mitigation or longitudinal control (Adaptative Cruise Control - ACC) systems. In this paper, we propose a method to perform generic, real time and robust road obstacles detection using stereovision. Our algorithm can detect any kind of obstacle on the road (vehicle, motorbike, pedestrian, etc.) even in the event of partial occlusions. The detection is performed whatever can be the longitudinal profile of the road (planar or not planar), at frame rate (25 Hz). A lane detection process is also performed to remove non-dangerous (i.e. out-of-lane) obstacles. Our approach is based on the construction and subsequent processing of the “v-disparity” image, which provides a suitable geometric representation of the road scene despite possible matching mistakes between the two images. We present the method, experimental results and a longitudinal control system showing that the approach is suitable in the automotive context. (Author/publisher)
Abstract