带匹配支持度的多源遥感影像SIFT匹配方法
A SIFT Image Match Method with Match-Support Measure for Multi-source Remotely Sensed Images
-
摘要: 通过将匹配支持度的相似性测度引入SIFT特征匹配算法,提出了一种能够应用于不同源遥感影像的自动匹配方法。首先,建立待匹配影像中特征点的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,挑选一定数量的距离最为接近的匹配点作为候选点;最后,分别计算候选匹配点间的匹配支持度,并通过松弛法剔除误匹配点以完成影像的自动匹配。实验结果表明,与传统的灰度匹配及经典的SIFT特征匹配相比,此算法可明显提高影像匹配的成功率和可靠性。Abstract: An automatic image match method based on SIFT features with match-support measure is presented for multi-source remotely sensed images.In order to adjust SIFT match algorithm for multi-source remote sensing images,the match-support measure is introduced for similarity measure.First,a SIFT feature descriptor is built and the points satisfying the minimum Euclidean distance of the candidate matched points between reference image and match image are selecud.Second,we calculate the match-support measure among these candidate matched points respectively.Finally,we employ the relaxation method to discard the false matched point pairs.A stereo of SPOT-5 HRG imagery and aerial image are selected and used for experiment.The empirical results are compared with the results of traditional grey correlation method and conventional SIFT feature match method.The results show that the SIFT feature match method with match-support measure is reliable and efficient for automatic multi-source remotely sensed image match.
下载:
