GAN-Based Target Shadow Background Generation for SAR Identification

The patent introduces a generative adversarial network (GAN) as an intermediary system between the SAR image input and target identification output. The GAN includes a generator that creates synthetic target shadow background images and a discriminator that evaluates these images, enabling the system to focus on target features rather than background features. This intermediary structure resolves the contradiction by providing a sophisticated feature extraction mechanism that improves identification accuracy despite increased system complexity.

The patent replaces traditional mechanical or algorithmic target identification methods with a neural network-based GAN system. The generator and discriminator networks use learned representations to automatically distinguish target features from background features, substituting conventional image processing techniques with machine learning-based approaches that achieve superior identification accuracy.

The patent employs preliminary action by pre-training the GAN model on extensive SAR image datasets before actual target identification tasks. The generator and discriminator are trained in advance to learn the complex relationship between targets and their shadow backgrounds, so that during actual operation, the pre-trained model can quickly generate accurate TSB images without requiring extensive computation during real-time processing.

The generator in the GAN creates synthetic copies of target shadow background images that mimic real SAR image characteristics. These generated TSB images serve as approximations or copies of what the actual target shadows would look like, allowing the discriminator to learn from these synthetic examples and improve target identification without requiring exhaustive analysis of every possible target configuration.