Research On Structured Compressed Sensing And Signal Processing Approaches For SAR/InSAR

Synthetic Aperture Radar(SAR)is a microwave imaging radar,which is capable of acquiring two-dimensional radar backscattered coefficients about an area of interest via active sensing and signal processing.By combining two or more SAR images,interferometric SAR(In SAR)extends the mapping ability of SAR from two-dimensional space to three-dimensional one.With increasing demands on SAR mapping ability in modern society,SAR and In SAR are faced with a number of difficulties and challenges.For SAR imaging,the data acquisition is conventionally based on the shannon-Nyquist sampling theorem.However,with the increasing demand on SAR image information like high resolution and full polarimetry,the conventional sampling theorem puts increasing pressures on analog-to-digital converter(ADC)sampling rate and data storage.For In SAR,the data acquisition and signal processing are based on the conventional principle of common-band interferometry.The principle requires that In SAR data should have the same resolution and observation mode.Otherwise,interferometric processing needs to take low resolution as the basis,which converts the interferometric data pair from different resolutions and imaing modes to the same resolution and imaging mode.However,this processing will unavoidablely decrease the interferogram quality,leading to some unwanted effects,like reduced resolution,increased noise level,and blurring artifacts.Modern advanced SARs,like Gaofen 3,usually have multiple resolution levels and imaging modes,so it is meaningful to study how to use SAR data acquired with different resolutions and imaging modes for interferometry.Based on in-deepth investigations on the above problems,this study proposes a type of structured compressed sensing(CS)model for characterizing these problems.Therefore,it bulids a fundation for solving the aforementioned SAR/In SAR issues via the CS approach to reduce SAR data sampling rate and enable In SAR with different resolutions and imaging modes.The main research contents and innovations of this study are listed as follows:1)A two-dimensional random sequence modulation-based structured CS model is proposed.The proposed CS model injects randomness into the measurement process via random sequence modulation,which follows the CS criterion of using random measurement to obtain a sensing matrix satisfying the restricted isometry property(RIP).Theoretical analyses show that the proposed model satisfies RIP condition with high probability.This property provides robust signal reocvery with respect to unideal sparsity of signals and meansurement noise.The proposed CS model provides a theoretical fundation for solving the SAR/In SAR problems in this study.2)A sub-Nyquist SAR scheme based on quadrature compressive sampling(Quad CS)is proposed for reducing SAR raw data sampling rate and size.The proposed Quad CS SAR scheme has a structured CS model that is mentioned above,and thus the RIP is satisfied and the recoverability of sparse SAR images is guaranteed.For fast imaging with compressed SAR data,a two-dimensional sparse imaging method is designed.Numerical simulation results show that the Quad CS SAR scheme can efficiently reduce SAR data sampling rate and provide a satisfactory reconstruction performance of SAR images.3)A CS-based InSAR method is proposed for the interferometry between two stripmap images acquired with different resolutions.The proposed method uses CS as the signal processing tool,and retrieve interferograms via sparse signal optimization by using the transform-domain sparsity of interferograms,which can obtain improved interferogram quality than the conventional common band filtering-based method.Theoretical analyses show that,due to the speckle effect in radar imaging,the sensing matrix of the proposed model is inherently random.The sensing matrix is a two-dimentional random sequence modulationbased structured sensing matrix as mentioned above which satisfies RIP with high probability,and thus the recoverability of interferograms is guaranteed.For fast recovery of interferograms,an algorithm with low computational complexity is derived.Simulation results show that the proposed approach can obtain interferometric phase with improved quality compared with the conventional method.4)A CS-based InSAR method is proposed for the interferometry between two images acquired with different imaging modes and resolutions.The proposed method extends the above-mentioned In SAR method using stripmap images with different resolutions to the case of multi-mode In SAR problems.Specifically,based on the In SAR principle and bandlimited acquisition function of various imaging modes,this study derives an interferometric relation for images pairs acquired in different modes by means of analyses from frequency domain to time domain and from point target to extended target.The derived model is a general one,which can be applied to multiple mode combinations,e.g.,staring spotlight-sliding spotlight and stripmap-TOPS.The proposed model is linear and ill-posed,for which CS is used for model inversion.Based the aforementioned structured CS model,the CS matrix of the proposed In SAR model can satisfy RIP with high probability,and thus the recoverability of interferograms is guaranteed.For dealing with the spectrum weighting effect introduced by different antenna patterns and focusing algorithms,a spectrum weight compensation method is provided based deramp processing.Real-data experiments and simulation results show that the proposed method can obtain interferometric phase with improved quality.