针对高分影像在获取过程中产生的几何形变导致匹配困难问题,本文设计了一种SIFT与互信息相结合的匹配方法,通过对五层十五级的同一地区不同时相GF-1/GF-2影像进行基于GPU的SIFT算子特征提取,并通过欧式距离进行特征点匹配,在欧式距离特征匹配的基础上,以互信息作为相似性测度,通过正向和反向的双向匹配方法得到匹配点对,然后利用RANSAC算法将由于光照等因素导致的错误匹配点对剔除.结果表明:通过SIFT提取特征点,结合欧式距离和互信息两种匹配算法,然后通过RANSAC算法将光照等因素导致的误匹配点剔除可以有效地提高匹配正确率和匹配精度.%In order to solve the problem of the geometric deformation caused by the high grade image in the ac-quisition process,a matching method of sift and mutual information is designed,and the SIFT operator feature extraction based on GPU is carried out by the GF-1/GF-2 image of different time phase in the same region of level five layer 15.By using the euclidean distance to match the feature points,on the basis of the euclide-an distance feature matching,using the mutual information as the similarity measure,the matching point pairs are obtained by the forward and reverse bidirectional matching method,and then the error matching point pairs caused by the illumination and other factors are eliminated by the RANSAC algorithm.The results show that the matching accuracy and matching precision can be improved effectively by using the SIFT to extract the fea-ture points,combining the two matching algorithms of Euclidean distance and mutual information,then elimi- nating the error matching points caused by illumination and other factors by RANSAC algorithm.
展开▼
