The Research On Road And Obstacle Detection Based On Stereo Vision

With the milestone leaps in computer computing power and the continuous improvement of sensor precision,more and more researchers have turned their attention to advanced driver assistance systems in order to bring better driving experience to drivers and reduce accidents..Compared with other active sensors,vision-based sensors can not only provide intuitive images,but also can be deployed on a large scale in applications.There is no problem of signal interference between each other and can provide a lot of information for decision-making.In addition,thanks to the improvement of GPU parallel computing technology,the real-time processing of image matching has become a reality,and the use of vision to detect obstacles has gradually become the mainstream of current research.Therefore,this paper detects roads and obstacles in traffic scenes based on binocular vision,and the main tasks completed are as follows:(1)Study the binocular vision model and the principle of stereo matching.Derive the binocular vision system model and its target detection principle through the extension of the camera imaging geometric model,analyze the basic implementation steps of the stereo matching algorithm used to calculate the parallax and the advantages and disadvantages of the typical algorithms,and the constraint criteria and similarity measures that affect the parallax results Functions are introduced.(2)Research the road surface extraction algorithm based on binocular vision.The algorithm first uses the SGM algorithm to obtain the disparity map,and uses the bicubic interpolation method to optimize the disparity map;then the disparity map is projected in the horizontal and vertical directions to generate the UV disparity map,and the road surface is in the V disparity map according to the RANSAC algorithm.The slash represented by in is fitted;finally,the cost image is introduced to correct the extraction result of the road surface.(3)Research on obstacle detection algorithms based on binocular vision.On the basis of road surface extraction,the road surface information and background information in the disparity map are removed to determine the target obstacle area;then the base point of the obstacle is calculated through the UV disparity map,and different constraints are established for obstacles of different scales to realize the traffic scene Accurate detection of obstacles in the target.(4)Test and analyze the proposed road and obstacle detection algorithm.By setting the comparison algorithm and the verification parameters,the comparison experiment is carried out on the standard data set,and the four pixel-level performance evaluation measures of the correct rate Q,the accuracy rate P,the recall rate R,and the comprehensive evaluation index F are used to evaluate the detection performance of the algorithm in this paper.Comprehensive evaluation and analysis.