Electromagnetic Deep Learning Technology for Radar Target Identification

We propose a new electromagnetic deep learning technology to radar target identification implemented via a novel spatial singularity expansion method (S-SEM). The proposed approach utilizes a recently-found radiation function that holds the spatial parameters of targets defined as the surface current and the geometrical details. Through EM machine learning, an estimation of these parameters is performed in a form of inverse problems for a single wire system. The estimated parameters, which are the S-SEM data, length and orientation, are validated and compared with numerical results obtained from the EM solver where an excellent agreement is achieved.