The Research Of Microwave Correlated Imaging Methods In Non-ideal Case

Microwave correlated imaging based on two-dimensional temporal-spatial stochastic radiation field and correlated processing method is a novel radar imaging technique whose resolution is not limited by the antenna aperture,and has the potential to achieve uninterrupted observation and imaging of targets in a fixed area.It opened up a new development direction for the current high-resolution radar staring imaging.Under ideal environment,the traditional correlated processing algorithms can achieve good imaging results.However,in practical engineering applications,due to the influence of non-ideal factors in the radar system,such as low SNR,weak randomness of the radiation field,and the model errors,the imaging quality of the existing algorithms is degraded.Therefore,this thesis mainly studies the correlated processing algorithm that can achieve better imaging results in non-ideal case.First of all,in this thesis,the specific process of microwave correlated imaging is analyzed in detail,and the expressions and correlations of echo signal,stochastic radiation field and scattering coefficients are obtained by formula derivation,and the signal model of microwave correlated imaging is established.Since the correlated imaging resolution depends on the randomness of the radiation field,this thesis introduces the effective rank as an evaluation index to measure the randomness of the radiation field.In order to further meet the requirement of randomness of the radiation field,this thesis uses a random frequency hopping signal as the transmit signal to construct the radiation field,which can increase the spatially-independent radiation field samples and improve the correlated imaging resolution.Then,compressed sensing theory and factor graph theory are introduced,which depends the foundation for the study of new correlated imaging methods.Secondly,to solve the problem that the poor imaging quality of some traditional correlated imaging algorithms in the case of low SNR or weak randomness of radiation field,and the correlated imaging algorithm based on Sparse Bayesian learning(SBL)has very high computational complexity and is difficult to real-time imaging,this thesis introduces a Sparse Bayesian learning algorithm based on approximate message passing with unitary transformation(UTAMP).The radiation field matrix is transformed into the product of the diagonal matrix and the unitary matrix by the unitary transformation,and the model of correlated imaging is transformed into a SBL model,and the joint probability density function is modeled as a factor graph model.Update the probability estimates of hidden variables and hyperparameters by performing the message passing on the factor graph,avoids the complicated matrix inversion operation in the SBL algorithm,which greatly reduces the computational complexity and improves the imaging efficiency.The simulation results under different system conditions show that the algorithms can still achieve ideal imaging results under the case of low SNR and weak randomness of radiation field.Finally,aiming at the problem of model mismatch caused by various error factors in practical engineering application and the effects of block-sparse target on imaging performance,this thesis studies the block-sparse target correlated imaging method under the model mismatch.Firstly,the imaging model is modified on the basis of analyzing various error factors,and the effects of the block-sparse target on the sparse reconstruction algorithm is verified by simulation.Then,this thesis improves the traditional total variation(TV)regularization,adding gradient information in the diagonal and anti-diagonal directions.In addition,this thesis uses the total least square(TLS)algorithm to solve the error matrix,weight ed total variation(WTV)regularization is added to constrain the optimization function,and an optimization problem suitable for block-sparse target imaging under model mismatch is proposed.Then,this problem is solved by augmented Lagrange multiplier method and alternating direction method.Finally,simulation results show that the proposed algorithm has good imaging performance on block-sparse target in the case of model mismatch.