一种基于邻域投票的异源光学影像SIFT匹配误差剔除方法

异源遥感影像匹配是高分影像处理中的重要环节与关键问题，但目前异源高分影像匹配精度有待提高。本文提出了一种基于邻域投票的异源光学影像SIFT匹配误差剔除方法，首先利用尺度不变性特征变换（SIFT）对特征点进行提取，随后基于邻域投票对匹配特征点进行二次约束，最后区分出待剔除误差大的匹配点，进而确定精确匹配点。为了评价本文方法的精度，分别对建筑物、道路、水体进行匹配研究，试验证实该方法可以提高上述3种地类的匹配精度，相比传统的SIFT方法平均提高了66%，同时有效地保持了结果的尺度不变性。

A Method of SIFT Matching Error Elimination for Heterogeneous Optical Images Based on Neighborhood Voting

Heterogeneous remote sensing image matching is an important part of the high score image processing and key issues,but the current heterogeneous high score image matching accuracy needs to be improved. In this paper, a method of SIFT matching error rejection for heterogeneous optical images based on neighborhood voting is proposed. First this paper uses the scale invariant feature transform (SIFT) to extract the feature points.Then,the secondary constraint is applied to the matching feature points based on the neighborhood voting.Finally,this paper distinguishes the error to be removed from the matching point,and determines the exact match point.In order to evaluate the accuracy of the method, it researches on the buildings, roads, water body matching respectively. Experiments show that the method can improve the matching accuracy of the above three kinds of land types,compared to the traditional SIFT method increased by an average of 66％, while effectively maintaining the scale of the scale invariance.