一种可靠的高分辨率光学卫星遥感影像匹配方法

A Reliable Matching Approach for Optical High Resolution Satellite Images

提出一种基于虚拟交点的高分辨率光学卫星遥感影像自动匹配方法。具体流程包括:(1)初始获取同名点,建立影像局部粗匹配模型;(2)利用同名点构建同名虚拟直线,通过同名虚拟直线形成同名虚拟交点集,并采用局部粗匹配模型进行约束;(3)对候选点集进行特征描述;(4)对特征点利用最小欧式距离准则提取初始同名点;(5)采用RANSAC算法和多项式拟合迭代法剔除误匹配点以获取最终的匹配结果。实验结果表明了本文通过虚拟交点提取同名点的算法,获取了更好的匹配效率和精度。

This paper presents a matching method for high resolution optical satellite images based on virtual corner.Firstly,initial matching are proposed to extract the homonymy points which is used to build local image transform model;secondly,virtual lines are built by the homonymy points,then virtual corners dataset are constructed on the basis of virtual lines with the constraint of local transform model;thirdly,features descriptors are performed;fourthly,the minimum Euclidean distance criterion is used to extract the initial homonymy points;at last,RANSAC and iterative polynomial error checker are imbedded to eliminate mismatching points.Experimental results verify that the presented algorithm with virtual points is more efficient on the quantity and accuracy of correspondence points.