Research On ISAR Imaging Algorithm Based On Deep Learning And Fourier Ptychographic Microscopy

As a high-resolution detection technology,Inverse Synthetic Aperture Radar(ISAR)imaging systems are widely used in military and civilian applications.In practical applications,the electromagnetic environment and target motion state of the radar signal are very complex,so the requirements for imaging accuracy,real-time and robustness of the algorithm are very high.Traditional imaging methods can no longer meet the standards of ISAR high-precision imaging.Various solutions combining deep learning and radar imaging algorithms such as data-driven and model-data-driven have been proposed.Since the radar operating environment and the motion of the target are usually very complex,how to display the characteristics of the maneuvering target stably even under low signal-to-noise ratio conditions is the focus of research in the field of deep learning ISAR imaging.Based on this,two deep learning-based high-resolution ISAR imaging algorithms are proposed in this paper.The main research contents of the article are as follows:First,a geometric model for ISAR imaging of maneuvering targets is established,and the Range Doppler(RD)imaging algorithm and the Fourier Ptychographic Microscopy(FPM)imaging algorithm are highlighted.These algorithms serve as pre-processing methods for the deep learning networks in this paper and are the subsequent support of this paper.Second,to address the problems of low imaging resolution and severe scattering of the traditional ISAR imaging algorithm for maneuvering targets,this paper proposes a highresolution reconstruction scheme for ISAR based on the U-Net network.The network is trained using FPM preprocessed data,and the final simulation shows that the model can obtain high-resolution ISAR imaging results of rotating targets.At the same time,the algorithm has good noise immunity and can clearly recover the contours of moving targets for subsequent target identification in noisy environments.Finally,an ISAR high-resolution reconstruction scheme combining Transformer and FPM is proposed to address the problem that the ISAR images recovered by the U-Netbased ISAR imaging network model is blurred under the low signal-to-noise ratio condition.The scheme adopts a local-global training strategy and also uses a self-attentive module for feature extraction of the input feature map,which makes the error of this network training smaller.The experimental results show that the algorithm can achieve high-resolution ISAR imaging and recover clear ISAR images even in the case of high noise.