A Nonlinear Inverse Scattering Approach towards RCS Modeling of Arbitrary Targets for Digital Twin Applications

Automatic Target Recognition (ATR) efforts for Synthetic Aperture Radar (SAR) applications are often challenged by the scarcity of diverse training data for relevant targets. Recently, digital twins have been employed as a means to augment measured SAR imagery for the purposes of ATR training [1]. Such approaches are based on the use of Computational Electromagnetic (CEM) techniques to generate accurate radar cross section (RCS) data (and subsequent SAR imagery) for targets of interest. Unfortunately these approaches lack scalability as they rely on appropriate digital models of the underlying targets. In an effort to promote a more efficient and scalable digital twin construction process, our work explores the use of inverse scattering for the purpose of digitizing arbitrary physical targets in a manner that is amenable to efficient and faithful RCS prediction for arbitrary aspect angles within the desired band of operation.