A Super-Resolution Algorithm for Extended Target Localization

We aim to localize extended targets (i.e. non-point targets) using stepped-frequency (SF) radar measurements from multiple transceivers. In particular, we consider a general signal model with directionally varying target reflectivity that can account for many types of extended target geometries. We localize targets under this signal model using a novel convex relaxation of a tensor atomic norm minimization problem that leverages discrete prolate spheroidal sequences (DPSS's). We test this algorithm on simulated data and show that our proposed method outperforms alternatives for our considered signal model.