SAR Image Generation Algorithm Based on Azimuth Partition Directed Generative Adversarial Network

Due to the insufficient of SAR images, the accuracy of automatic target recognition (ATR) often fails to meet the requirements of practical applications. In this paper, we propose an azimuth partition directed generative adversarial network (APDGAN) to generate SAR images. The proposed algorithm divides the azimuths into several partitions, and constructs an integrated condition that combines the category and azimuth partition information as the input of the network. The proposed algorithm realizes joint control of the category and azimuth to generate SAR images. We evaluate the generated images on the moving and stationary target acquisition and recognition (MSTAR) dataset. The effectiveness of the proposed algorithm is verified through recognition accuracy and Fréchet inception distance (FID) values.