基于SIFT特征的多源遥感影像自动匹配方法  

Auto-matching between multi-source remote sensing images base on SIFT features

作　　者：李然[1] 张云生[2] 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079 [2]武汉大学测绘遥感国家重点实验室,武汉430079

出　　处：《测绘科学》2011年第3期8-10,18,共4页Science of Surveying and Mapping

基　　金：国家973计划资助项目(2006CB701302);国家创新研究群体科学基金资助项目(40721001)

摘　　要：本文提出一种基于SIFT特征的不同源遥感影像自动匹配方法。首先利用Harris算子结合SIFT特征提取影像上均匀分布的特征点,建立高维SIFT特征描述符;然后以待匹配点与参考点间的欧氏距离为相似性测度,实施两种不同源遥感影像的特征匹配;最后将SIFT特征匹配结果作为初始值,采用将搜索范围扩大到尺度空间后的改进最小二乘影像匹配方法对SIFT特征匹配获得的同名像点进行精匹配。经对同一地区SPOT-5 HRG全色影像与航空摄影影像的匹配试验,取得了较好的结果。In this paper,an automatic image matching algorithm base on SIFT features was presented for misaligned multi-source remote sensing images.In the algorithm,the feature points were extracted by the scale-space extrema detection,and interesting value of Harris operator which was used to make the distribution of feature points evenly.Then the SIFT feature description was created and base on the matching rule of the Euclidean distance similarity measurement,the homonymy points between different source remote sensing images were obtained.At last,using the result from SIFT feature matching as initial value,the homonymous points could be forwarded for precise location by Least Square Matching.The experiment data was SPOT-5HRG panchromatic image and aerial image,and the experiment result was satisfying.

关 键 词：SIFT特征 多源遥感影像 影像匹配 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]