Fast ISAR Image Prediction for Targets with Coating Defects Through Deep Learning

The radar absorbing material (RAM) coating defects existing on the surface of stealth aircraft have important effects on their electromagnetic (EM) scattering characteristics. Considering the complexity and randomness of defects and the large electrical dimension of the platform, it is necessary to analyze their scattering characteristics through a series of complex and time-consuming processes including geometric modeling, meshing, EM simulation, and ISAR imaging, thus cannot meet the requirement of real-time analysis. To address this issue, this paper proposed a novel end-to-end deep neural network (DNN) based on residual U-net, which can perform time-efficient ISAR image prediction of a target with random coating defects from the input 2D geometric map of it. Compared to the method of shooting and bouncing ray (SBR) simulation and range-Doppler (R-D) ISAR imaging, the proposed DNN can accelerate the speed by three orders while ensuring a relative error lower than 1%. Numerical results are exhibited to verify the accuracy and efficiency of the proposed method.