Fig. 1. Principle of the GAN

Fig. 2. Network architecture of the SDSE-GAN

Fig. 3. Internal structures of the generator and discriminator of the SDSE-GAN

Fig. 4. Some images of the dataset expanded by the SDSE-GAN

Fig. 5. Common types of exterior surface damage of spacecrafts identified in image samples

Fig. 6. Network architecture of YOLOv5

Fig. 7. Inspection results of model inference

Fig. 8. Surface damage dataset of the NEU-CLS steel

Fig. 9. Comparison of the training metric parameters before and after data augmentation