高分辨率遥感图像配准控制点均匀化算法  

A Control Point Uniformization Algorithm for High-resolution Remote Sensing Image Registration

作　　者：马旭燕 袁媛[2] 汪承义[2] 陈静波[2] 贺东旭[2] 

机构地区：[1]青海省第二测绘院,青海西宁810001 [2]中国科学院遥感与数字地球研究所,北京100101

出　　处：《遥感信息》2016年第3期24-30,共7页Remote Sensing Information

基　　金：国家自然科学基金(41301499)

摘　　要：针对多源遥感图像普遍具有数据量大、辐射差异大等特征,而现有的图像配准算法无法直接应用于遥感图像自动配准处理中的问题,综合考虑控制点的密度和分布,提出了一种高分辨率遥感图像自动配准算法。首先,将待配准图像和参考图像降采样到单机可以处理的大小,利用尺度不变特征变换(SIFT)算法建立降采样图像间的初始匹配;其次,将原始待配准图像按照网格分割为子图像,并利用初始匹配找到每幅子图像在参考图像上的对应子图像;再次,利用SIFT和极大稳定区域(MSER)特征点的空间互补性,在每一对子图像上提取大量特征点;最后,利用随机采样一致性(RANSAC)算法剔除误配后,采用基于最大团问题的贪心法进行控制点均匀化处理,进一步剔除冗余的控制点。与现有的基于SIFT特征和基于灰度的遥感图像配准算法相比,本算法在配准精度和控制点的分布均匀度等方面具有优越性。To solve the problem of large volume remote sensing image registration,considering the density and distribution of the control points,a novel high-resolution remote sensing automatic registration algorithm is proposed.Firstly,down sample the two images to the size that can be processed by a single computer,and scale-invariant feature transform(SIFT)method is used to find the initial matches.Secondly,based on the initial correspondences,the input image is cut into sub-images with a uniform grid,and for each sub-image,its corresponding sub-image in the reference image is clipped according to the initial matches.Thirdly,SIFT and maximally stable extremal region(MSER)features are extracted separately from each pair of subimage for their fine spatial complementarity.Finally,after the removal of false matches by RANSAC technique,redundant control points are also eliminated based on the greedy method of the maximum clique problem.Compared to existing SIFT-based and intensity-based remote sensing image registration method,the proposed algorithm has better performance in terms of registration accuracy and control point distribution.

关 键 词：高分辨率遥感影像配准 降采样预匹配 SIFT MSER 控制点均匀化 

分 类 号：P237[天文地球—摄影测量与遥感]