Array SAR3D Imaging Method Based On Compressed Sensing

As a newly come out radar system, array SAR (ASAR)3D-imageing system wasdeveloped based on conventional SAR2D-imageing system. The radar images of riseand fall terrains (like alpine valleys and city block) reconstructed by ASAR, whichovercomes the shadow effect and space fuzzy problem, are with high resolution.3D-RDalgorithm and3D-BP algorithm are much used methods to reconstruct images, whichare conventional image reconstruction method and the resolution is limited by Rayleighlimit. As the resolution of radar images reconstructed by conventional method is hard toenhance, super-resolution3D image reconstruct methods are proposed recently whichcan obtain images with higher resolution relative to conventional methods. There havetwo main technology paths of image reconstruction method, spectral estimationmethods and compressive sensing, that can efficiently suppress the side-lobe and reducethe width of main-lobe. In this thesis,3D image reconstruction methods based oncompressive sensing will be discussed in detail. The main content of this thesis are:1. Three working modes of linear array SAR system are briefly introduced and thedistance history of radar echo is discussed. Two conventional3D image reconstructionmethods are presented and the shortcomings of the conventional methods are discussed.Compressive sensing theory is briefly stated and the sparsity of target in3D space hasbeen analyzed which show that the application of compressive sensing in radar systemis feasible. Then, the linear observation model of the linear array3D SAR system isderived in detail.2.3D image reconstruction method based on sparsity driven autofocus (SDA) isproposed by applying SDA in SAR3D imaging system which convert the imagereconstruction problem to constrained optimization problem. The derivation of thelinear observation model with phase error has been discussed in detail and the SDAobservation model is given. Simulations have been given which show that3D imagereconstruction method based on SDA is feasible and the images obtained by SDAmethod have the superiority of side-lobe suppression and focusing effect.3. Weighted Sparsity Driven Autofocus (WSDA) method is proposed and the cost function of WSDA adopts weighted L1norm constraint. The solving processes of theconstrained optimization problem of WSDA are presented in detail. To demonstrate thesuperiority of WSDA in side-lobe suppression and focusing effect compared with SDA,simulation results are given.4. Threshold Orthogonal Matching Pursuit (TOMP) algorithm is proposed and thesuperiority of the TOMP compared with OMP is demonstrated by simulations. Thelinear observation model with real signal form of linear array SAR3D imaging systemhas been derived in detail. Based on the real form observation model, BCS and GradientPursuit methods are induced to3D imaging system, the feasibility of theabove-mentioned two methods are proved by simulations. Several sparsityreconstruction methods are discussed to give the guiding principle when apply sparsityreconstruction methods in radar imaging area. CoSaMP and SP methods are notappropriate for radar imaging and the reasons have been discussed. Four sparsityreconstruction methods which are suitable for radar imaging are presented, BCS, GP,TOMP and WSDA. The influences of SNR on the four methods are discussed.Comparison analysis of super-resolution ability and efficiency of the four methods aredemonstrated by simulations. A conclusion of good and bad points of the four methodsis given which can provide a guiding principle of how to choose and apply sparsityreconstruction methods in SAR imaging area.