Image Data Augmentation for SAR Automatic Target Recognition Using TripleGAN

Synthetic aperture radar (SAR) images can be observed in all weather conditions. In addition, SAR automatic target recognition (ATR) is an important technology in the field of remote sensing image analysis. However, it is expensive to acquire SAR images, which limits the construction of additional datasets. In this paper, we propose a method to enhance ATR performance of the corresponding SAR images based on triple-generative adversarial networks (GANs). We added a classifier to learn together so that generator converges into the real data distribution. Experiments on the MSTAR dataset confirmed that the proposed model is applicable through the classification model.