Research On InSAR Complex Image Registration And Interferogram Filtering

Interferometric Synthetic Aperture Radar(In SAR)is an active three-dimensional remote sensing mapping technology which has the advantages of large observation range,high measurement accuracy,all-day and all-weather observation on the ground.It has significant application value in digital elevation reconstruction,surface deformation monitoring,glacier movement research and so on.In order to obtain reliable In SAR measurement results,it is necessary to process the interferometric data with high accuracy.However,due to the decoherence effect between SAR data and the periodicity of interferometric phase,In SAR data processing is challenging.In order to improve the accuracy of In SAR measurement results,this paper focuses on the first two steps of In SAR data processing,namely SAR complex image registration and interferogram filtering.The main contents of this paper are as follows:(1)Aiming at the problems that the real and complex cross-correlation function registration algorithm of SAR complex image can not select high coherence control points independently and the adaptability of single measure function is poor,a high-precision two-stage registration algorithm is proposed.Firstly,the robust real cross-correlation function registration algorithm is used to get the coarse interferogram and coherence coefficient image.Then,the selection of high coherence control points and the selection of fine registration method are carried out to avoid the low coherence of control points selected by cross-correlation function registration or improper selection of registration measure function.The experimental results show that,compared with the real and complex cross-correlation function registration algorithm and the spectrum ratio registration algorithm,the proposed algorithm has higher coherence and better phase continuity.(2)Aiming at the problems of inaccurate terrain phase extraction and unadaptive residual phase filtering in the interferogram phase compensation filtering algorithm,a phase compensation adaptive filtering algorithm is proposed.The improved algorithm can effectively improve the accuracy of terrain contour phase extraction by prefiltering the interferogram and smoothing the spectrum of the prefiltered interferogram with Goldstein method.In the residual phase filtering,the adaptive multi-look filtering based on the pixel coherence coefficient can effectively filter the noise and reduce the loss of phase information.The experimental results show that,compared with some commonly used interferogram filtering algorithms,the proposed algorithm can effectively eliminate the phase noise and avoid the phase loss.(3)Aiming at the problem that the vector Euclidean distance can not effectively measure the pixel similarity of interferogram when the phase noise is large,a robust nonlocal mean filtering algorithm is proposed.Firstly,the phase compensation adaptive filtering algorithm is used to pre-filter the interferogram,which can keep the phase structure and reduce the phase noise to a lower level.Then,the high reliability pixel similarity is calculated based on Euclidean distance of phase vector,which can effectively improve the performance of Nonlocal means filtering.The experimental results of simulated and measured interferograms show that the proposed algorithm has the best filtering performance in simulated interferograms compared with the three comparison algorithms,while in measured interferograms,the proposed algorithm produces a certain phase structure ambiguity,and a more effective filtering algorithm still needs to be studied.