【摘要】 针对月球车立体视觉系统对匹配速度和准确度要求较高的特点,提出了一种基于特征约束的区域匹配算法。该算法用多尺度小波变换实现边缘提取和降噪,采用多特征匹配、双向匹配、部分二次采样、选取图像子集等技术保证匹配的速度和精度。最后对匹配结果进行滤波,进一步提高匹配的准确性。更多还原

【Abstract】 A matching algorithm was brought forward based on the high matching speed and high matching accuracy required of lunar rover vision system in this paper. With a feature-constrained area-based matching method as the central part, the algorithm used multi-scale wavelet transform to detect edges and reduce noise, and utilized such technologies as multi-feature matching, two-direction matching, partial sub-sampling and image window selection to ensure the speed and accuracy of matching. Finally the disparity map was filtered to further improve the accuracy of matching.更多还原