A Multi-modal Fusion and Mesh-free RCS Prediction Method Based on PointNet++

Mesh-free method has been a hotspot for calculating electromagnetic scattering characteristics of non-cooperative targets, as the inability of traditional methods to obtain model information of non-cooperative targets and to compute the scattered field by the computational electromagnetic method without meshing. The fitting capabilities of deep learning have quickly gained attention in this area due to their potential to overcome these challenges. This study proposes a novel deep learning framework for predicting the Radar Cross Section (RCS) of PEC objects made of random complex structures under specific conditions. This framework utilizes point cloud and even images as the input, aiming to integrate simulation information such as frequency, angle, and polarization to construct a dataset, and directly predicts the RCS of targets made of PEC material under these specific conditions.