Research And Performance Evaluation Of MIMO-SAR Near Field High Precision Imaging Methods

High resolution millimeter wave radar systems require high cost and complexity,and adopting a compromise between MIMO and SAR technology can effectively reduce system cost and complexity while achieving imaging resolution.In near-field imaging applications,traditional imaging algorithms often rely on approximation of far-field signal models,resulting in insufficient imaging accuracy;Deep learning methods can achieve high-precision imaging,but they rely on a large amount of datadriven,lack forward physical models,and are difficult to converge.Based on the above shortcomings,this article derives a joint radar physical model driven and data driven deep network imaging method and its sparse imaging method to achieve high-precision imaging of MIMO-SAR near-field targets.The main research work is summarized as follows:Firstly,based on the radar far-field signal model,the traditional equivalent element midpoint approximation method has been improved,and an accurate MIMO-SAR nearfield signal model has been introduced.Introduced several traditional SAR imaging algorithms,extended them to MIMO-SAR imaging through MIMO array error correction method,provided several performance evaluation indicators,conducted simulation experiments and traditional algorithm imaging performance analysis,and elaborated on the corresponding deep learning theoretical knowledge,providing theoretical basis,dataset support,and imaging performance comparison for the deep network imaging method derived in this project.Secondly,a joint radar physical model driven and data driven deep network imaging architecture SAR-Net is proposed.On the basis of solving the inverse problem of deep learning radar imaging,a radar near-field signal model is introduced,and the solving steps are expanded into a deep network architecture based on the conjugate gradient algorithm.Each layer of the architecture is a fusion of CNN denoising blocks and image reconstruction blocks.And derive the SAR-Net 3D imaging method,which compensates the imaging results of different distance planes based on phase difference and overlays them to achieve 3D imaging without the need for repeated model training.The simulation experimental results show that SAR-Net can achieve high-precision image reconstruction of 2D and 3D targets with a small number of training samples.Finally,to verify the feasibility of SAR-Net in practical applications,an imaging resolution analysis,signal waveform design,and overall system design were conducted,and a MIMO-SAR near-field signal acquisition system was built.In order to reduce the time cost and storage cost of the system data acquisition process and SAR-Net computing process,the SAR-Net sparse imaging method is derived based on the compressed sensing theory and tested through simulation experiments.Finally,imaging experiments were conducted using measured and open-source data collected by the system.The results showed that SAR-Net and its sparse imaging method have high practical feasibility and universality,and the SAR-Net sparse imaging method can effectively reduce the computational and storage costs of the data acquisition and imaging processes.