机器人和驾驶中的计算机视觉.pptVIP

Proceedings of the 2001 IEEEInternational Conference on Robotics AutomationSeoul, Korea ? May 21-26, 2001Computer Vision Based Object Detection and Recognition for Vehicle DrivingLi-Chen FuComputer Science Information Engineering,National Taiwan UniversityTaipei 106,TaiwanE-mail : lichen@.twCheng-Yi LiuComputer Science Information Engineering,National Taiwan UniversityTaipei 106,TaiwanE-mail : r7526008@.twABSTRACTa system can detect the existence of an obstacle and tell ifit may affect the safety of the vehicle driving, then thediver can be asked to be more careful of such a conditionso that a possible traffic accident may be avoided. This isespecially important when the driver is drunk or when thevehicle is traveling at a very high speed, and then manylives, including the driver and pedestrians, could be saved.On the other hand, the system may also facilitate thesubsequent design of an automatic vehicle guiding systemwhereby the driver only needs to specify the destination orto choose the strategy for cruising when he is inside avehicle for driving.Applying computer technology to vehicle driving hasbeen studied for many years. Most of the studies focusedon autonomous vehicle driving in a simplified environmentlike freeways, or life independent systems like GPS. Ageneral case where kind of unexpectedly fatal conditionsmay occur which we are driving in an urban area, however,has not been considered. In this paper, we proposed asystem to satisfy the basic criteria for such a generaldriving assistance. We detect the obstacles on the groundin front of the vehicle we are driving, and then classifythem into three predefined categories: pedestrians,vehicles, and others.Related WorkIn the proposed system, we exploit a simplifiedstereovision system to detect the obstacles instantly. Afterthat, to search and track pedestrians and vehicles,different methods are used for adapting to theirheterogeneity. For the sake of implementation, we alsopropose a method to decide the max