LONG DISTANCE VISION SENSOR FOR DRIVER ASSISTANCE

Intelligent vehicle research to date has made great progress toward autonomous driving or driver assistance. Integrated systems for on-road vehicles, which include road following, lane departure, obstacles detection and avoidance have been demonstrated. However, one of the weakest points of current automated vehicles, is that there is no reliable system to detect obstacles on the road surface at longer distances. Classically systems for obstacle detection and recognition use stereovision sensors. But, using stereovision images at this distance, the detection of road boundaries and small obstacles is a difficult problem since information is not available because the resolution is too weak. In this paper we propose a new stereovision sensor capable to obtain images with sufficient resolution at long distance, in order to be able to discriminate between obstacle, lane and objects. This stereo sensor just uses a single-lens camera, a prismatic mirror placed in front of the camera, and mirrors on each side and parallel to the prismatic mirror. Because the prismatic mirror is rotated around its vertical edge the field of view of the stereoscope can be oriented toward the region of interest. The development of this catadioptric sensor is a part of a project, name ‘Ravioli', which deals with the huge problem of Intelligent Vehicle. The French acronym ‘Ravioli’ stands for Radar Vision Orientable Lidar.