一种月球车视觉系统的匹配算法

A Matching Algorithm for Lunar Rover Vision System

对采用双目视觉来实现月球车自主导航提出了一种快速匹配的方法。首先,对相机的内外参数进行精确地标定,通过两相机的外参数对图像进行核线纠正,生成消除上下视差的核线;然后,在核线图像上,采用相关系数法进行由少到多的粗匹配,对匹配结果进行多重的检验,接着用最小二乘方法进行亚像素的精确匹配;最后,在匹配的像点间构建Delaunay三角网,建立两图像重叠区域的匹配关系,实现稠密匹配。对该方法进行了多组实验,实验结果表明:该方法可以快速、可靠地实现稠密的图像匹配。

This paper put forwards a matching algorithm for lunar rover vision system that will use double cameras tO navigate. Firstly, the elements of interior and exterior orientations of cameras are calibrated, and then the original images are matched to epipolar images without vertical parallax by elements of exterior orientation between two stations, and then points on epipolar images are matched increasingly by correlation coefficient algorithm, and the matched points are checked by many rules. Secondly points are matched accurately by least square match algorithm. Finally, Delaunay triangulated network between matched points is established and the relation algorithm of overlap regions between epipolar images is obtained. Some experiments are performed by the method in the paper. The experiment results indicate that the method can finish matching pictures of lunar rover quickly and reliably.