Research On Microwave Staring Correlated Imaging Of Low-Rank And Large Scene

Microwave staring correlated imaging(MSCI)method based on temporal-spatial stochastic radiation field which resolution is not limited by the antenna aperture and can realize staring of fixed area is a novel radar imaging system that has proposed in recent years,it has very important research value.Microwave staring correlation imaging emits space-time two-dimensional stochastic radiation field through multiple transmitting antennas of the antenna array,then radiation field interacts with the imaging region target,and the receiver receives the echo containing the target information.By associating echo and spatio-temporal two-dimensional random radiation fields,high-resolution imaging of the target is obtained.This dissertation focuses on the research of microwave staring correlation imaging algorithm in low-rank and large-scale scenes.Firstly,the low-rank scene imaging algorithm is studied.The mathematical model of microwave staring correlation imaging is introduced,and the existing microwave staring correlation imaging algorithm and its main problems are analyzed combined with the model,then proposes two correlation imagingalgorithms for low-rank scenes:1)The low rank constraint and total variation regularization constraint are added in the process of correlation processing between echo and stochastic radiation field for the restoration of the low rank scene target.The low rank constraint uses the prior information of the low rank of the low rank scene,can reveal the structural redundancy information hidden in the data.Total variation regularization method has a good effect on image denoising and texture edge retention.Combining the low-rank constraint and the global variational regularization constraint can simultaneously utilize the global and local information of the scene.For low rank constraints,a non convex function which is closer to the rank function than the kernel norm is proposed to approximate the rank function.By adjusting the parameters,the whole solution problem is a convex problem.Simulation results show that the proposed method is effective for low rank scene targets.2)In the imaging process,there is often a lot of noise.For sparse noise,the low-rank sparse decomposition of the scene target is performed during the recovery process,and the scene matrix is decomposed into a low-rank matrix containing a scene target structure information and a sparse noise matrix.Combining the total variation regularization method and optimizing low-rank and sparse parts similtaneously during correlation processing,the scene target can be effectively recovered from the observation matrix with sparse noise.The simulation of the effects of sparse noise on the target shows that the method has a good recovery effect.Secondly,the algorithm of microwave staring correlation imaging in large scenes is studied.In order to solve the problem of large scale imaging equation,the imaging equation is preprocessed by multi grid processing method.The large-scale imaging equation is decomposed into different size grid space to process,and the high frequency error in imaging square is eliminated by smoothing iteration on the same layer,and between different layers,the iteration eliminates the low frequency error in the imaging equation.In the coarsest mesh,the Least square QR-factorization(LSQR)algorithm which is more effective than other conjugate gradient algorithms to solve the ill conditioned large equation is used to obtain more accurate initial solutions.Simulation results show that the proposed method can improve the imaging results and shorten the imaging time.