Based On Structural Features SAR And Optical Image Registration Algorithm Of Research

As two common types of remote sensing data,synthetic aperture radar and optical images are widely used in image fusion,change detection and other fields.Image matching is the basic step and key link of these applications,but due to different imaging mechanism,there are large nonlinear intensity,geometric difference and speckle noise between SAR and optical images,it is difficult to mine the common features of the image directly by using the existing matching methods,and then can not recognize enough matching points.In order to solve the above mentioned problems,a new method of SAR and optical image matching based on structural features is proposed in this paper,which takes SAR and optical image as research objects and focuses on intensity invariant structural feature extraction and description,to improve SAR and optical image matching performance.The main contents of this paper are as follows:(1)Modified Uniform Nonlinear Diffusion-based Harris feature detector is proposed to solve the problem of uneven distribution of feature points and low repetition rate in SAR and optical images.In contrast to the traditional UND-Harris operator,which extracts feature points directly from the original image,this operator extracts feature points from the Phase Congruency structure graph.By incorporating Phase Congruency information into the UND-Harris operator,the operator is more resistant to the nonlinear intensity difference between SAR and optical images,and can extract feature points with uniform distribution and high repetition rate.(2)Phase Congruency Order-based Local Structural descriptor is proposed to solve the problem of poor robustness and salience of SAR and optical image feature descriptors.In order to enhance the saliency of the descriptor,the descriptor divides the feature regions into different sub-regions according to the Order of the PC values of the pixels,which is used to extract the structural features of the image.Compared with the traditional method which adopts the sequence of image intensity values directly,this method can effectively improve the discrimination of descriptors and the invariance of intensity difference,and improve the robustness of descriptors,the descriptor uses adaptive spatial structure to encode image structure features.Compared with the traditional method,the proposed descriptor is more robust to the geometric differences of images.To sum up,this thesis focuses on two key techniques of SAR and optical image feature point extraction and description.Experimental results show that the proposed method is more robust to nonlinear intensity and geometric difference.Compared with the existing methods,the proposed method can significantly improve the performance of SAR and optical image matching.