Single-Channel Circular Synthetic Aperture Radar Moving Target Detection Based on Deep Neural Network and Autofocus Algorithm

Abstract

Phase errors on account of target motions induce the smearing of images acquired by circular synthetic aperture radar (CSAR). Motivated by this phenomenon, a novel moving target detection method for single-channel CSAR based on deep neural network and autofocus algorithm is proposed in this paper. In this method, the phase gradient autofocus (PGA) algorithm is first applied to correct phase errors. Then, the deep residual network (ResNet) is utilized to learn the information concerning the variation of SAR patches before and after PGA. Considering the limitation of labeled real data in practice, the combination of simulated moving targets and measured background clutter is employed to train the network. As is verified by experiment results, the proposed detection method overperforms the current autofocus-based moving target detection approach.