中文摘要：

考虑到特征点在特征维度上的有序性,通过排序配对的方法减少错误匹配点,降低错误匹配率。将RANSAC算法应用到SIFT算法中,获得初步的匹配点对;对匹配点分别进行横、纵坐标的排序,对照剔除序号不一致点;计算两幅图像的质心,将各匹配点到质心的距离排序,剔除序号不一致点;将剩余的匹配点作为最终的正确匹配点。实验结果表明,在尺度变换、噪声干扰、旋转变换和错切变换条件下,改进算法的正确匹配率在一定程度上有所提高。

英文摘要：

Considering the order of feature points on the characteristic dimension,error matching points were reduced using sor-ting and matching method to reduce the error matching rate.RANSAC was applied to SIFT and the initial matching points were obtained.The matching points were sorted according to the abscissa and ordinate respectively to remove the inconsistent points. The centroids of the two images were computed respectively and the distance between matching points and centroid was sorted to remove the inconsistent points.The rest of the matching points were taken as the final correct matching points.Experimental re-sults show that under conditions of scale changing,noise interference,image rotation and shear transformation,the correct matching rate of the improved algorithm is enhanced to some extent.