Research Of Obstacle Detection Method Of Cyber Car Based On Stereo Vision

In recent years,the focus of the automotive industry is moving toward intelligence.For intelligent vehicles,the perception of the environment is the basis for the precise positioning and path planning of the vehicle.Vision can provide rich scene information.Therefore,in this paper,method for detect obstacles based on stereo vision is proposed.Firstly,this paper gives an overview of the camera imaging model of the stereo vision system,including the three coordinate systems and the transformation between the coordinate systems.In the ideal stereo vision system,the line-aligned left and right views are obtained after stereo correction,using the semi-global matching algorithm to get disparity map,and the edge filtering is utilized to get optimized disparity map reflecting the depth information based on the left view-guided.The map can provide accurate 3D information for subsequent traffic scene obstacle detect.Secondly,this paper illustrated the U-V disparity map theory and construction method,as well as the common plane in the U-V disparity map.The ground plane will form a straight line in the V-disparity map,the plane equation of the road surface is obtained by detecting the straight line,so that the ground can be filtered out,and the part above the ground is the obstacle area.Obstacle detection based on V-disparity maps sometimes results in incomplete ground filtration and the exclusion of obstacle regions.The stixels algorithm combined with Udisparity map is proposed to detect obstacles.The disparity values of roads are calculated according to the principle of V-disparity,road scores and obstacle scores are calculated for each pixel,cost images are generated,The intersection of road and obstacles each column is obtained by dynamic programming.The obstacle height is determined by the membership function,then the obstacle area is determined,the U-parallax operation is performed on the obstacle area,the column coordinate of the obstacle is determined by Hough line detection,and the position of the obstacle is finally determined.Finally,the hardware and software configuration needed for the stereo vision system was established to verify the effectiveness and real-time performance of the algorithm in detecting obstacles ahead of the vehicle.