景象匹配算法及其适应性分析

Scene matching algorithm and its adaptability analysis

提出一种基于区域特征的景象匹配算法。其基本方法是先提取基准图和实时图的稳定极值区域,并计算出区域的平均灰度、方差、仿射不变矩以及纹理信息熵,构造相似性度量函数,然后计算匹配区域对的Hausdorff距离决定是否是对应区域。大量的模拟实验表明,该算法的匹配时间能达到0.138 s,正确匹配率达到92.67%,不仅能满足景象匹配的要求,且大大提高了匹配的时间与正确率。对其适应性进行了详细的分析,结果表明该算法对旋转、尺度、噪声、灰度变化的适应性较好。

A stereo scene matching algorithm based on regions feature is proposed. Its basic method is extracting the regions feature of the reference image and real image and computing the gray value, shape parameter, texture information and variance of region to design the similarity measurement function firstly. Then the Hausdorff distance of the region pair to decide whether corresponding regions is computed. Plenty simulation experiments show that the matching time may achieve 0.138 s, the best correct matching rate is 92.67%. The results show the algorithm not only meets the requirements of the scene matching, but improves the matching time and correct rate greatly. Finally, its adaptability is analyzed, and the results show the algorithm has good adaptability for rotation, scale, noise and gray changes.