Super Resolution Imaging Of Linear Array SAR

Linear array SAR,which could obtain the two dimensional resolution through composing the fictitious planar antennas and achieve the third dimensional high resolution by the pulse compression,could realize the 3D imaging.It can reconstruct 3D scene comprehensively,and has been become a research focus so far.However,the difficulty of imaging is that the distributions of antenna elements are restricted by the structure and payload of the wings,so,due to the decrease of the length of array in cross-track direction,traditional algorithms would bring about image quality problems.Aimed at this problem,this thesis,based on the multiple input multiple output technique,studies the super-resolution imaging algorithms of both uniform linear array and non-uniform linear array,which is researched as follows:(1)When the flight altitude is low,the resolution in cross-track direction is approximate to that in along-track direction.In order to increase the 2D resolution,a fast 2D-ESPRIT super-resolution imaging algorithm,based on the uniform linear array,is proposed.Firstly,the conventional ESPRIT algorithm is expanded from one to two dimensionality.Then,the L-shape array is utilized to replace the whole virtual flat antenna elements,and the algorithm is combined with the RD algorithm to calculate the scattering coefficients,which could reduce the computational complexity.(2)When the virtual array could only be the non-uniform linear array in some platform of the special structures,most of the spectral estimation algorithms wouldn't achieve a satisfactory resolution due to the structure of array.In order to improve the imaging quality of the non-uniform linear array,the compressed sensing algorithm based on the redundant frame is proposed.It could improve the resolution in cross-track direction by transforming the structure of the measure matrix.(3)For the imaging procedure in LASAR system,CS imaging is an ergodic process on each height and along-track cell.Hence,the computation cost and the memory consumption will be significantly large.The truncated singular value decomposition is proposed to compress the measure matrix.It could reduce the computational complexity,while the imaging quality would not be debased.