Fast Generation of RCS Data Based on Scattering Mechanism Analysis and infoGAN

Abstract

The deep neural network can be used in fast generation of radar cross section (RCS) data. The training dataset is important in the construction deep neural network. To make sure that the deep neural network can learn the accurate relationship between a target's parameters and RCS data, the scattering mechanism is suggested to be considered in preparing the training dataset. In this work, we propose a new method to prepare the RCS training dataset based on scattering mechanism analysis. The geometric model of the target is split into several facets, and ray tracing method is applied on such model. Then the facets are correlated by the clustered rays and scattering field. The key geometric parameters in the target can be obtained by analyzing the related scattering mechanism which contributes the most to the RCS data. Based on the sweeping of these key paremeters, the RCS dataset can be constructed. Moreover, we use such dataset to train infoGAN. Numerical results validate such method.