Research On High-precision Image Registration And Phase Filtering Of Interferometric Synthetic Aperture Radar

Interferometric Synthetic Aperture Radar(InSAR)technology has great research significance in many fields such as military monitoring,disaster prevention and mitigation,ecological environment construction and so on.Compared with traditional remote sensing mapping technology,InSAR technology has great advantages,occupies an important position in the field of national economic construction and national security,and has become a hotspot of international research.This paper mainly studies two key technologies in the whole process: one is InSAR complex image registration,which needs to achieve 0.1 pixel registration accuracy.The processing accuracy of this technology directly affects the complexity of subsequent image data processing;the other is InSAR phase filtering.The phase noise formed on the interferometric phase image will affect the real value of the interferometric phase,making the subsequent phase.The unwrapping process produces errors,so it is necessary to eliminate noise by phase filtering,so as to obtain more accurate interference phase.The main work of this paper includes the following aspects:1.For incoherent InSAR image pairs that may occur,a rough registration method based on SIFT framework for InSAR image pairs is proposed.In the rough registration stage,the algorithm is divided into four steps based on SIFT framework: scale space construction stage,using rolling guide filter to construct scale space,can effectively remove the inherent coherent speckle noise while retaining the edge structure of the image;feature point detection stage,using ROEWA operator combined with Harris-Laplace detection operator to accurately extract high position.The invariant feature points with low precision and low error rate;the description based on M-SURF is used in the construction stage of descriptor;the feature search based on KNN algorithm is used in the feature search and matching stage to determine the set of points to be registered,and the feature matching based on RANSAC algorithm is used to obtain the global transformation matrix.2.Aiming at the high accuracy requirement of complex image registration,an InSAR image registration method based on Delaunay triangulation is proposed.In the stage of precise registration,the algorithm divides sub-regions based on Delaunay triangulation,and solves the estimation of the pixel difference in sub-regions according to the analysis model,and corrects the image according to the calculated value to remove the sub-pixel deviation between the primary and secondary images,so as to complete the process of accurate correction of the local areas of the image.3.For the additive noise model of interferometric phase,an InSAR phase filtering method based on non-linear diffusion filtering is implemented,which solves the problem that the statistical filtering method does not take into account the influence of terrain on the selection of parameters,thus overcoming the disadvantage that the traditional method destroys the original phase image while filtering.