基于SURF和改进RANSAC算法的图像自适应匹配

摘要

移动机器人在环境中精确定位依赖于快速准确的图像匹配,传统的SURF匹配算法存在大量的错误匹配,不能满足实际要求.从匹配时间和匹配正确率2个方面对其改进,在特征匹配过程中,通过双向FLANN搜索算法和预匹配筛选出大量的误匹配点,然后与S-RANSAC算法结合,优化匹配结果,得到正确的匹配结果.%Accurate positioning of mobile robots in the environment depends on fast and accurate image matching.Due to the existence of a large number of errors, the actual requirements cannot be met by using traditional SURF matching algorithm.The SURF algorithm was improved in this study from two aspects of matching time and matching accuracy.In the feature matching process, a large number of false matching points were screened out through the two-way FLANN search algorithm and pre-matching, and then combined with the S-RANSAC algorithm, the matching results were optimized.Finally, the correct matching results were obtained.