Spaceborne Interferometric Synthetic Aperture Radar Algorithm Of A Number Of Key Technologies

Interferometric Synthetic Aperture Radar(InSAR) is a new developed microwave remote sensing technique.It is known by its unique advantage of all-weather,large-region monitoring and high-resolution imaging for the measurement of 3D DEM and earth surface deformation.InSAR has been applied widely in topography mapping,vegetation classification,3D reconstruction of artificial buildings and geodynamics related research such as surface subsidence,volcano and iceberg activity etc.3D DEM and earth surface deformation can be retrieved from two single look complex(SLC) image with minor difference after a processing chain of registration,filtering, interference and phase unwrapping etc.This dissertation mainly focuses on two key steps in InSAR data processing:SLC images registration and phase unwrapping.In addition,a method which is used to simulate SLC images with different quality is proposed.Firstly,the history developing of InSAR,especially the retrieval of DEM by InSAR,is reviewed,and the current situation of InSAR research in domestic and abroad is depicted briefly in this dissertation.The theory and conventional algorithms of InSAR processing steps are summarized and discussed.The development of InSAR DEM retrieval and the current situation of InSAR research in domestic and abroad are reviewed firstly.And then the theory and conventional algorithms of InSAR processing steps are summarized and discussed.A coarse-to-fine registration framework,based on the mean square difference(MSD) function,is introduced in this dissertation.Many windows whose size is 256×256 are chosen equally and the coarse offsets of windows in slave image are determined when the MSD function of every window reaches its minimum.After filtering out the nonoverlap spectrum in azimuth,hundreds of windows whose size are 64×64 and disperse equally on slave image are used to measure subpixel shifts.Taking the similar step as coarse registration processing,fine registration determines subpixel shifts by moving windows with 0.1 pixel interval.The centers of those moving windows with higher similarity are chosen as control points for next step registration.The corresponding relation between the two images is computed from 2D-polynomial transformation model of those selected control points.After the slave image is interpolated with Raise Cosine kernel,offset of each subpixel in slave image relative to the corresponding pixels in master image is obtained. According to the offsets,the entire slave image is resampled and registered with the master image.The experimental results,showing clear fringes,verify the validity the registration procedure.A new approach based on Ant Colony Optimization(ACO) is presented to show good feasibility for phase unwrapping of the DEM retrieval.This approach seeks the shortest path linking all residues via ACO algorithm and then divides it into segments of balanced residues,in order to avoid isolated regions which can not be unwrapped.Due to the optimization strategy to establish the branch cuts,the unwrapping error can be reduced significantly.Experiments of simulated dataset and real ENVISAT data are employed to test the proposed approach.The unwrapping results indicate that ACO algorithm is an optional compromise strategy between preferable unwrapping precision and less time consuming.Aiming at interferograms with large size and dense residues,two path-following unwrapping algorithms which are suitable for different distribution of the dense residues are presented:the method of multi-patches connecting opposite residues and the method of center expanded searching opposite residues.The idea of the former is to segment the interferogram into many balanced patches according to the minimum distance principle of opposite residues,whereas the latter method is to search and connect the nearest neighbor opposite residues beginning from the center region of the interferogram to its boundary.The former is appropriate for interferograms whose residues disperse in many patches and the latter suits for interferograms whose residues are well-distributed as a whole.They improve the unwrapping accuracy for they avoid repetitive connection of residues and longer branch cuts.The algorithms are employed to ENVISAT datasets,clearly demonstrating their tangible improvement on unwrapping accuracy over other path-following unwrapping algorithms.Experimental results show that the two methods,which are applicable for interferograms with dense residues dispersing differently, have the advantages in execution time and precision. A SLC image simulation method is presented in this dissertation.Based on the SAR imaging principle and scattering model of two-scale rough surface, the amplitude and phase of the SLC image is computed from the real DEM under the orbit parameters of ENVISAT-ASAR.Several sets of SLC images with different shadow are simulated by changing the differently scaled DEM and then interferograms with different residues are generated from the simulated SLC images,which can be used to verify the InSAR algorithm and analyze the influence of InSAR shadow.