Improvement of SAR Target Classification Using GAN-based Data Augmentation and Wavelet Transformation

Synthetic aperture radar (SAR) is a powerful tool inremote sensing. Unlike optical image devices, SAR can observe target regions regardless of weather conditions, such as clouds, fog, and darkness. In this article, we consider the SAR target classification problems when available SAR images having target labels are limited. To improve the classification performance, we propose alearning technique combining data augmentation usinggenerative adversarial network (GAN) models and wavelet transformation. We conduct experiments to investigate the improvement of the proposed learning technique with the SAR images from the moving and stationary target acquisition and recognition data. From our experiment results, the proposed learning technique combining GAN-based data augmentation and wavelet transformation has shown greater improvement in SAR image classification when the available learning data is scarce.