Change Detection For Polarimetric SAR Image Based On Scattering Characteristics And Low Rank And Sparse Decomposition

Synthetic Aperture Radar(SAR)which is a high resolution microwave radar,possessing the characteristic of all-weather,all-time capabilities,has become one of the important means of obtaining target information.With the development of polarimetic SAR(POLSAR),scattering information of target detected by radar is becoming more and more plentiful,SAR image is more widely used in the field of classification,detection and recognition.The change detection of SAR image is a comparison of two or more SAR images acquired at different times in the same area,and the process of changing information of objects is obtained according to the difference between images.Change detection of POLSAR image is an important component of SAR image processing,and has wide application prospect in agriculture,urban change disaster assessment etc.Compared to conventional SAR images mainly containing intensity information,the POLSAR image contains a large amount of information,such as intensity information,scattering power information,phase information,polarization scattering information,etc.It is a difficult problem for POLSAR image detection how to use these information in multi-temporal POLSAR image to improve the completeness of difference in difference image.The paper focuses on how to improve the completeness of difference imformation in difference image by using polarimetric characteristics in multi-tempral POLSAR data.And how to make use of the sparse property of the multi-temporal remote sensing images,we have finished two main tasks as following :(1)A method for POLSAR image change detection based on scattering power characteristics and low rank sparse decomposition is proposed.Firstly,the mutli-temporal POLSAR data are decomposed by decomposition methods to obtain polarimetric characteristics,and effective characteristics,which be selected from the above polarimetric characteristics by Log-ratio method,are used to construct an input image sequence of increasingly changed region.Secondly,the input image sequence is decomposed by the sparse low rank decomposition method to obtain the sparse image sequence that expresses the change region.Difference image is obtained by merging the sparse image sequence.The low rank sparse method used in the proposed method not only considers the neighborhood information in single image,but also considers the difference information between the images.Finally,the method of weighted mean fusion fuses the sparse image sequence.The experimental results that the completeness of difference information in difference image is improved and noise is restrained.(2)A method of change detection for POLSAR image based on double difference image fusion is proposed.Firstly,intensity characteristics which are extracted from POLSAR data,is input images of the two phases.Difference image 1 is obtained by low-rank and sparse decomposition.And the scattering power characteristics which are obtained by using the method of Freeman decomposition from POLSAR data,is multi-temporal input of POLSAR images.Difference image 2 is obtained by low-rank and sparse decomposition.Then the difference image is fused between difference image 1 and difference image 2.The experimental results show that this method makes full use of the complementarity between features,and effectively improves the accuracy of change detection.