SAR Data Expansiontechnology Based on Improved DCGAN
Jiangxin Qiu · 2023
Due to the difficulty of obtaining SAR images for non-cooperative targets, the application and development of SAR target recognition technology is greatly limited. This paper proposes a method based on improved DCGAN for SAR image data augmentation. By making appropriate improvements to the DCGAN network, the quality of the SAR images generated by the network under small sample conditions is effectively improved. Finally, the experimental results of the sample data set expansion effect are verified by YOLO v5. The experiments show that compared with two typical GAN models, DCGAN and WGAN, the SAR sample data set expanded by the improved DCGAN not only has high image quality, but also can effectively improve the recognition effect of the recognition network.