Multi-Frame SAR Image Super-Resolution Reconstruction Based On Nonlinear Kalman Filters

In the applying process,high-resolution synthetic aperture radar(SAR)images can be used to extract more information of ground objects.However,owning to the factors in the process of SAR imaging,such as imaging method,system hardware and atmospheric disturbance,the resolution of SAR images is below to the theoretical level and cannot meet the common application needs.As a result,this paper takes how to improve the resolution and quality of SAR images as the research object,takes digital image processing technology and the basic theory of SAR images as the research basis,takes data post-processing as the research mode,and carries out the research on the application of non-linear Kalman filter in the process of superresolution reconstruction of multi-frame SAR images,with the following main research work.(1)The theory of two unscented Kalman filter(UKF)-based multi-frame SAR image super-resolution reconstruction algorithms,importance sampling unscented Kalman filter(ISUKF)algorithm and adaptive importance sampling unscented Kalman filter(AISUKF)algorithm,are investigated.And the feasibility and effectiveness of the nonlinear Kalman filter for multi-frame SAR image superresolution reconstruction are validated by using simulated SAR image series.In addition,the peak signal-to-noise-ratio(PSNR),mean structural similarity index measure(MSSIM),and feature similarity index measure(FSIM)are used to quantitatively assess the quality of reconstructed SAR images.The ISUKF algorithm and AISUKF algorithm are comprehensively evaluated according to the numerical change curves of the three objective evaluation indexes and subjective vision of the reconstructed images.(2)Aiming at the reconstruction image shortages of the ISUKF and AISUKF algorithms,a novel multi-frame SAR image super-resolution reconstruction algorithm termed importance sampling cubature Kalman filter(ISCKF)are been proposed in this thesis.Firstly,the existing super-resolution reconstruction system model is improved by establishing the function relation between the equivalent number of looks(ENL)of low-resolution images and the parameter in the reconstruction system model.Combined with cubature Kalman filter(CKF),the ISCKF algorithm are been proposed.The numerical results of the reconstructed SAR images show that the DAMRF dynamic optimization model effectively enhances the robustness of the super-resolution reconstruction system.Since noise still remains in the image reconstructed by the ISCKF algorithm,the FT-WNNM model are derived based on the noise model of the image and the probability density function of the noise.A method to solve the noise reduction model is developed: alternating direction method of multipliers based on the low rank model(ADMM-LR).At the same time,the ADMM-LR algorithm is used as a post-processing of the ISCKF algorithm.Thus the multi-frame SAR image super-resolution reconstruction algorithm,ISCKF+ADMMLR,are been proposed.The numerical results and visualization of the reconstructed SAR images show that the ISCKF+ADMM-LR algorithm has strong advantages in term of noise suppression,reconstruction performance,detail recovery and preservation.