应用于多源SAR图像匹配的级联SIFT算法

Cascade SIFT Matching Method for Multi-Source SAR Images

针对多源SAR(Synthetic Aperture Radar)图像几何精处理需要大量离散控制点的问题,文章中提出一种级联SIFT(Scale Invariant Feature Transform)匹配算法.首先,采用大尺度自适应各向异性高斯SIFT(Adapted Anisotropic Gaussian-SIFT,AAG-SIFT)算法进行图像粗配准,大尺度AAG尺度空间可以在模糊不稳定局部纹理干扰的同时,保持图像的结构信息,提高算法的鲁棒性;其次,级联一种局部SIFT匹配算法,在粗配准后图像间进行局部匹配,避免不相关区域内重复纹理对特征匹配的影响;最后,通过尺度和旋转等先验条件筛选匹配点对,保证匹配结果的准确性.对比实验表明,级联SIFT处理可以增加提取同名点的数量和空间分布质量,而且匹配点定位准确.

This paper presents a cascade SIFT matching method to get large numbers of discrete distributed matches for accurate procedure of multi-source SAR images. Firstly,to obtain more robust matching result,a large scale adapted anisotropic Gaussian scale-invariant feature transform（ AAG-SIFT） method is proposed. It is constructed based on large scale AAG scale space,w here unstable local textures are blurred,but the structures remain unaffected. Secondly,a local SIFT matching method is recommended w hich extracts SIFT features from the coarse registered images. And then the features are matched in the local regions to avoid the interferences of the repeated textures from uncorrelated regions. Finally,the priori condition of scale and rotation consistency is used to filter matches,guaranteeing the accuracy of the matching results. Compared w ith traditional matching methods,the proposed method increases the number and distribution quality of the matches,and the location of the matches is accurate.