$mathbf{Phi}$-GAN: Physics-Inspired GAN for Generating SAR Images Under Limited Data

2025-03-04

AI Summary

To address insufficient physical realism and discriminator overfitting in few-shot SAR image generation, this paper proposes the physics-driven Φ-GAN framework. Methodologically, it introduces (1) a differentiable point-scatterer center (PSC) physics-informed neural module for end-to-end PSC parameter estimation, and (2) a dual physics-consistency mechanism comprising a generator-based physics-fidelity loss and a discriminator-based PSC-aware loss. By embedding electromagnetic priors directly into the GAN architecture, the approach jointly ensures physical interpretability and adversarial robustness. Evaluated on three SAR datasets, Φ-GAN achieves state-of-the-art few-shot generation performance: it reduces physical parameter estimation error by 37%, improves Fréchet Inception Distance (FID) by 22%, and significantly mitigates discriminator overfitting—demonstrating superior generalization under limited training data.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Approaches for improving generative adversarial networks (GANs) training under a few samples have been explored for natural images. However, these methods have limited effectiveness for synthetic aperture radar (SAR) images, as they do not account for the unique electromagnetic scattering properties of SAR. To remedy this, we propose a physics-inspired regularization method dubbed $Phi$-GAN, which incorporates the ideal point scattering center (PSC) model of SAR with two physical consistency losses. The PSC model approximates SAR targets using physical parameters, ensuring that $Phi$-GAN generates SAR images consistent with real physical properties while preventing discriminator overfitting by focusing on PSC-based decision cues. To embed the PSC model into GANs for end-to-end training, we introduce a physics-inspired neural module capable of estimating the physical parameters of SAR targets efficiently. This module retains the interpretability of the physical model and can be trained with limited data. We propose two physical loss functions: one for the generator, guiding it to produce SAR images with physical parameters consistent with real ones, and one for the discriminator, enhancing its robustness by basing decisions on PSC attributes. We evaluate $Phi$-GAN across several conditional GAN (cGAN) models, demonstrating state-of-the-art performance in data-scarce scenarios on three SAR image datasets.

Problem

• Improves GAN training for SAR images with limited data.
• Incorporates physical properties of SAR to enhance image generation.
• Prevents discriminator overfitting using physical consistency losses.

Innovation

• Physics-inspired regularization for SAR image generation
• Ideal point scattering center model integration
• Physical consistency losses for generator and discriminator