Research On SAR Radio Frequency Interference Mitigation Method Using Multidimensional Analysis

Compared with the optical sensors,synthetic aperture radar(SAR)has the full-time and all-weather imaging characteristics with a wide range of military and civil applications,such as terrain observation,disaster rescue,electromagnetic situational awareness and precision guidance.However,the changeable electromagnetic environment and aliasing spectrum resources make the large-bandwidth SAR system susceptible to radio frequency interference(RFI),which detracts from the SAR imaging quality and hinders the subsequent interpretation of images.Due to the limitation of RFI on information perception capability of SAR system,it is significant that substantial efforts should be dedicated to the research on SAR RFI mitigation method.There are abundant multi-dimensional characteristics of the observeted targets in the SAR data,such as the feature in the spatial domain,time domain,frequency domain,time-frequency(TF)domain,statistical domain,and image domain.However,the most existing RFI detection and mitigation researches only utilize the partial information in the single dimission of SAR echo data,resulting in the problems of low model accuracy and poor robustness.Meanwhile,the most existing RFI mitigation methods needs to refine the hyperparameters in the RFI mitigation model,which lacks the self-adaptiveness.Aiming at the challenges in the SAR RFI mitigation researches,this dissertation focuses on analyzing the multi-dimensional feature difference between RFI and target echo signal,and makes a profound study on the SAR RFI mitigation methods in the statistical domain,TF domain,and range-azimuth domain.It has significant theoretical value and wide application prospects with the support of national natural science foundation of china and the projects in science and technology commission.The main contributions of this dissertation are presented as follows:1.Firstly,this dissertation introduces the operation principles of SAR system,and deduces the SAR imaging algorithm of different working modes based on the SAR echo representation model.Then,some parametrical mathematical models of SAR RFI are established in the time domain,frequency domain,TF domain,statistical domain,and image domain.And the multi-dimensional characteristic differences between RFI and target echo signal are analyzed based on the measured SAR data,which laid the theoretical foundation for subsequent RFI detection and mitigation research.Finally,two RFI detection methods are introduced based on the TF skewness and TF kurtosis,utilizing the feature difference of TF structure and statistical distribution between RFI and target echo signal.The interference detection results from these methods can provide significant prior information for the subsequent RFI mitigation research.Meawhile,the effectiveness of the proposed RFI detection methods is verified based on the simulated and measured SAR datasets.2.Aiming at the problems of low self-adaptive level and poor robustness in the SAR RFI mitigation research,this dissertation introduced a new RFI mitigation scheme based on the variable Bayesian theory(VBIM).Firstly,based on the TF low rank characteristic of RFI and the Laplace distribution assumption of the target echo signal,a low-rank matrix factorization(LRMF)model with the L1 norm is constructed to transfer the RFI mitigation problem as an optimization process.Furthermore,to avoid the potential overfitting problem of deterministic model and improve the generalization ability of RFI reconstruction model,we extend it under the framework of Bayesian theory and formulate a hierarchical Bayesian complete posterior model based on the reasonable statistical representation of model parameters.Then,the variable Bayesian inference method is introduced to adaptively estimate the hyper-parameters of Bayesian posterior inference model,and the variational low bound is utilized to measure the iterative state of the RFI reconstruction procedure.Finally,the interference mitigation results based on both simulated and measured SAR datasets fully verify the effectiveness and robustness of the proposed VBIM algorithm.3.Aming at the challenges of low model accuracy and poor robustness in the SAR RFI mitigation research,a novel RFI mitigation algorithm for SAR data based on TF low rank and dual sparsity(LRDS)is proposed in this dissertation.Firstly,the short time Fourier transform(STFT)is utilized to obtain TF spectrogram of SAR echo signal,and the low rank and sparse properties of RFI are conducted through the eigenvalue and amplitude analysis of spectrograms.Combined with the sparsity assumption of the target echo signal in TF domain,a new RFI reconstruction model joint the low rank and dual sparsity is established.Then,the RFI reconstruction model is transformed into three sub optimization proplems,including low rank matrix approximation,RFI reconstruction and sparse recovery of target echo signal.Meanwhile,the bilateral random projection strategy and soft threshold mapping function are utilized to optimize the RFI reconstruction model alternately,and the recovery loss of target echo signal is further reduced through the interference cancellation strategy.Finally,RFI mitigation experiments on both simulated and measured datasets verify the effectiveness and robustness of the proposed LRDS algorithm.4.In order to mitigate the RFI in the complex electromagnetic environment,this dissertation proposes a novel RFI mitigation algorithm based on the range-azimuth constraints joint low rank and sparsity properties(RAC-LRS).Due to the high coupling features between the complex RFI and the target echo signal in pulse-based representation results of the SAR echo signal,the most existing RFI mitigation algorithms which process the SAR data pulse by pulse have the problem of model mismatch on this status.Firstly,to mitigate the RFI in complex electromagnetic environment,the range-azimuth joint representation model of the SAR data is constructed utilizing the azimuth information.And the SAR data matrix is preprocessed into several blocks to improve the execution efficiency of the interference mitigation algorithm.Then,the low rank and sparsity properties of RFI and target echo signal are introduced based on the qualitative and quantitative sub-block data analysis,and the RFI reconstruction model based on the low rank and dual sparse characteristics in range-azimuth domain is constructed.In view of the aggregation characteristic of RFI in the range-frequency azimuth-time domain,an improved RFI reconstruction model is formed based on the range-azimuth constraint joint low rank and sparse characteristics.Furthermore,the RFI reconstruction model is decomposed into two sub optimization problems,the low rank RFI estimation and sparse recovery of target echo signal,which is updated alternately until the precise reconstruction of RFI.Finally,the effectiveness and robustness of the proposed RAC-LRS algorithm are demonstrated based on its application on simulated and measured SAR data.