结合几何约束与尺度特征的光学与SAR影像配准  

An optical and SAR images matching method with geometric constraint and multi-scale feature congruency

作　　者：张一帆 李佳[1] 张亚 吴俊辉 李志强 ZHANG Yifan;LI Jia;ZHANG Ya;WU Junhui;LI Zhiqiang(School of Geosciences and Info-Physics,Central South University,Changsha 410083,China)

机构地区：[1]中南大学地球科学与信息物理学院,湖南长沙410083

出　　处：《海洋测绘》2024年第6期71-76,共6页Hydrographic Surveying and Charting

基　　金：国家自然科学基金(42374053);国家重点研发计划(2022YFB3903602);湖南省自然科学基金(2023JJ30656)。

摘　　要：现有光学与SAR影像配准方法受非刚性几何畸变、非线性辐射差异、不同分布模型的噪声等影响,精度还有较大提升空间。对此,提出一种结合几何约束与多尺度特征一致性(OSG-MFC)的光学与SAR影像配准方法。首先利用索贝尔(Sobel)与高斯拉普拉斯(LOG)算子提取光学特征点,基于指数加权平均比(ROEWA)与哈里斯(Harris)算子提取SAR特征点;然后对光学与SAR特征点进行几何约束,提取具有一致特征表达的连接点对,与规则格网点进行分级构建;最后建立相位一致性金字塔,构建多尺度匹配模板,实现光学与SAR的高精度配准。实验表明新方法绝对精度(RMSE)比目前4种主流方法提升20%~40%,正确匹配率(CMR)和RMSE分别可以达到39.5%、0.7个像素,为光学与SAR影像配准提供了一种新思路。To solve the problems of nonlinear radiation difference and non-rigid geometric distortion in optical and synthetic aperture radar(SAR)image matching,we proposed an optical and SAR image matching method with geometric constraint and multi-scale feature congruency(OSG-MFC).Firstly,two key-points positioning methods are proposed that conform to the corresponding image patterns and noise distribution.Specifically,the Sobel and Laplacian of Gaussian(LOG),and ratio of exponentially weighted averages(ROEWA)and Harris operator are used to extract optical and SAR images feature points.In addition,regular grid points are selected and initial correspondences are fitted based on the geo-transformation.The geometric constraint is applied to select the initial tie-points of optical and SAR images with feature expression congruency.During the matching process,a phase congruency(PC)pyramid is established based on the scale space theory,and a multi-scale feature congruency matching template is constructed based on the PC pyramid.Then a sliding template search is performed to locate the peak on the similarity map and to determine the precise matches.Matching experiments conducted on the registered optical and SAR images dataset show that the proposed OSG-MFC method improved the average matching absolute error(RMSE)by 20%~40%,achieved correct matching ratio(CMR)and RMSE to 39.5%,0.7 pixels,respectively.

关 键 词：遥感影像配准 光学与SAR影像匹配 特征提取 几何约束 相位一致性 

分 类 号：P237[天文地球—摄影测量与遥感]