SAR Automatic Target Recognition Based on Deep Convolutional Neural Networks

Abstract

Deep convolutional neural networks (CNN) have recently proven extremely competitive in challenging visible light image and speech recognition tasks. The goal of the present study is to explore the application of automatically learned convolutional network features to radar target recognition. Specifically, a two-stage convolutional-pooling network architecture is designed and error back-propagation algorithm with momentum acceleration strategy is used to learn the network weights in a supervised fashion. The effectiveness of the proposed method is assessed by SAR image classification tasks on the standard benchmark of MSTAR (Moving and Stationary Target Acquisition and recognition) database. Our experiments show the presented method has achieved encouraging results with a correct recognition rate of 95.64% for three classes of targets and 92.86% for ten classes of targets.