Research Of Key Algorithm Of Dense Point Map Generation Based On Binocular Vision

The dense point map of road scene has important applications in automatic driving,road planning and other fields.In recent years,with the development of computer vision technology,binocular vision technology has great advantages in obtaining complete and accurate dense point map.Using binocular vision to obtain dense point map has two key steps:visual localization and dense matching.The matching algorithm of sequence images is the main factor affecting the accuracy of visual localization,and the quality of disparity map obtained by stereo matching algorithm plays an important role in dense matching result.Based on the principle,two algorithms include combining optical flow with feature point matching algorithm and improved semi global matching using pyramid image and texture information constraints are researched as the following two aspects:1)Aiming at the problems of uneven distribution and low accuracy of optical flow tracking points which caused by camera irregular motion,a combining optical flow with feature point matching algorithm is researched.Firstly,matching triangles based on initial optical flow tracking points are constructed.Then,feature points are extracted by Difference of Gaussian algorithm,which descriptor computed using scale and rotation parameters calculated by matched triangles.Finally,the initial position of matching point are computed by angular intersection method,and final matching points are acquired.Two typical examples of sequential images are presented to prove the matching effectiveness,the experimental results show that the proposed algorithm can obtain more even tracking points,and matching points are used to estimate image pose have a good effects.2)Aiming at the problems of large computation time,memory overhead and poor edge preservation in disparity map obtained by semi-global matching(SGM),a modified semi global matching algorithm using pyramid image and texture information constraints is researched.In the stereo matching stage,the disparity map is computed using semi-global matching by layer after building the pyramids images,and an dilation-erosion algorithm for disparity graphs,which takes into account the image texture information,is introduced to constrain the range of parallax search,increase the number of effective pixels at the edge of the parallax map and reduce the memory overhead and computation time required for the algorithm.In the post processing stage of disparity image,the edge information of disparity image is protected by weighted median filtering algorithm.Two sets of KITTY stereo images are selected to test the matching effect.The experimental results show that the disparity map obtained by proposed method has a good accuracy,and the edge characteristics of the disparity are maintained well.The algorithm is computationally efficient and has relatively low memory overhead.3)Applying the two algorithms to the road dense point mapping experiment which use two sets of cvlib dataset,the results show that the proposed method has good effect in both whole and local mapping result,and the dense point map has good completeness and high accuracy.