A Fast RCS Prediction Method for 3D Dielectric Target Based on Machine Learning

In this paper, a method based on machine learning is proposed to accelerate the computation of radar cross section (RCS) for three-dimensional dielectric targets. Firstly, to construct a compact dataset, the input is changed from traditional mesh data to point cloud, and the traditional equivalent source distribution is expressed as inherent feature parameters (IFPs) with the help of characteristic mode theory. Subsequently, machine learning techniques are introduced to construct a neural network model for training the IFPs with reference to the PointNet neural network. Experiments show that the proposed method can accurately predict the RCS while significantly reducing the memory overhead and runtime compared with the method of moments (MoM), and achieves a fast and accurate end-to-end mapping from a target's structural attributes to its electromagnetic properties.