Research On Multi-Angle Polsar Image Registration And Change Detection Method

Polarimetric Synthetic Aperture Radar(PolSAR)is an active microwave sensor that can obtain the scattering information of ground objects in multiple polarization states and can provide rich information for the interpretation of ground objects.And compared with single-channel SAR,it can be widely used in military and civilian fields because of the better analysis and extraction of the polarization scattering characteristics.With the development of remote sensing technology,more and more PolSAR data can be used,so multi-temporal,multi-angle PolSAR image change detection has become a key technology in the field of PolSAR image processing,which can be applied to land planning and management,urban planning,environmental impact,natural disaster monitoring and so on.PolSAR image registration is a pre-processing step of change detection,and its accuracy directly affects the quality of subsequent change detection results.The research content of this paper is on multi-angle PolSAR image registration and change detection.The main work includes the following three aspects:First,the scattering characteristics of multi-angle urban ground objects are analyzed studied,Based on the characteristics of PolSAR images,typical urban target samples are selected and the characteristics of urban targets under various representation matrices are analyzed;Based on the target scattering theory,the scattering mechanism of urban ground objects are studied;Based on the target decomposition,the composition of the scattering mechanism of urban ground objects and the effects of different imaging angles on the scattering mechanism of the target are studied.Second,to solve the problem that the existing algorithms cannot make full use of the polarization information in the image,a PolSAR image registration algorithm based on stable feature point extraction and multi-feature fusion is proposed.By combining the polarization scattering components to calculate the gradient of pixels,the accuracy of feature point extraction in the registration process is effectively improved.In the SIFT framework,full variational model filtering is used to replace the traditional Gaussian filtering to construct the scale space,suppress the influence of coherent speckle noise on feature extraction,effectively retain the detailed information in the image,accurate gradient information are obtained and stable feature points are extracted.Experimental results show that the registration method in this paper is suitable for the registration of PolSAR images,and it has better registration performance than traditional methods for PolSAR images with complex scenes.Third,aiming at the problem that there are a large number of false change points based on the pixel-level and feature-level change detection results,and the target-level change detection process is cumbersome,a PolSAR image change detection algorithm based on multi-level information fusion is proposed.Based on the characteristics of rich polarization information and texture information in the PolSAR image of urban area,the extraction of the change area based on the fusion of pixel level and feature level is realized.Based on the super-pixel segmentation method,urban feature target image blocks are obtained,and the multi-feature fusion is used to extract the accurate change information of the candidate area to realize the multi-level information fusion PolSAR image change information extraction.Experimental results show that the change detection method in this paper has higher detection accuracy and lower false alarm rate and missed detection rate,and is more suitable for change detection of multi-angle PolSAR images.