SAR image classification via capsule networks

SAR image classification is considered as one of the most important subjects in Automatic Target Recognition (ATR). Therefore, identifying the correct class of targets has a significant importance to take a decision. On this subject, deep learning techniques, especially the convolutional neural networks (CNNs), have improved the performance for the problem of SAR images classification due to its powerful perspective of feature learning and reasoning. Yet, CNNs generally need a huge amount of data for training and do not accurately manage the transformations in the input data. Fortunately, the new machine learning approach that is recently proposed Capsule Networks (CapsNets) aims to overcome the drawbacks of CNNs. Specifically, the method proposed adopts and incorporates CapsNet for the SAR image classification problem by designing an improved framework which achieve better classification accuracy of our problem and performs the classification of SAR images. Results obtained while experimenting on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset exhibit the effectiveness of the adopted framework. Our results illustrate that the adopted method overcome successfully CNNs for SAR image classification.