3-D Parametric Metal Targets Inversion with Supervised Descent Method

In this article, we introduce an inversion method utilizing the supervised descent method (SDM) for reconstructing 3-D parametric metal targets. Trained on a set of simple targets, the SDM learns the mapping relationship between target characteristics and scattering data. This training process integrates prior knowledge into the optimization, effectively mitigating the ill-posedness of inverse scattering problems (ISPs). We demonstrate the application of this method to complicated targets, verifying its capability to predict accurate descent directions. The efficacy of this approach is validated through several examples, demonstrating its low training demands and high accuracy.