An optical and SAR images matching method with geometric constraint and multi-scale feature congruency
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.
remote sensing images registrationoptical and SAR images matchingfeature extractiongeometric constraintphase congruency
万方数据
维普
