基于特征点局部特征值剔除误匹配特征点算法

为了解决经典的特征点匹配算法SIFT采用比率测试得到的匹配特征点集中存在大量误匹配，且对数量和准确度无法兼顾的情况，提出了基于特征点局部特征值剔除误匹配特征点算法。该算法以高阈值比率测试得到的结果为粗剔除匹配点集，基于三角形相似性原理，从该特征点集中筛选出3个匹配正确的特征点对，利用其分别在基准图像和实测图像中构建局部直角坐标系，根据匹配的特征点对在相似局部坐标系下局部特征值的相似度剔除误匹配特征点，实现精剔除。实验结果表明，本文算法可以有效的剔除SIFT算法匹配结果中的误匹配，同时，与低比率（0.6）测试匹配结果比较，准确度较高，降低了匹配正确的特征点被误剔除的概率。可见本文算法可有效的剔除误匹配特征点，获得准确度高的匹配点集。

Elimination of Mismatched Feature Points based on Feature Point Local Eigenvaluesx

In orders to solve the problem that the classical feature point matching algorithm SIFT adopts the ratio test to obtain a large number of mismatches between the matching feature points set , and the number and accuracy cannot be considered together , Elimination of mismatched feature points based on feature point local eigenvalues was proposed . The algorithm was used to get the high threshold ratio test to obtain the rough matching point set. Based on the principle of triangle similarity , three pairs of matching feature points are selected from the feature points , and the local Cartesian coordinate system is constructed in the reference image and the measured image respectively . According to the similarity between local eigenvalues of matching feature point pairs to reject mismatched feature points for achieving accurate rejection. The results show that The algorithm can effectively eliminate the mismatches of the SIFT algorithm matching result . At the same time , compared with the low ratio (0.6 )test matching result , the accuracy is higher , which reduces the probability that the correct matching feature points is rejected by mistake . It is concluded that The algorithm can effectively eliminate the mismatches and get an accurate set of points.