Answers are grounded in this paper and cite their sources.
Recently, computer-aided design models and electromagnetic simulations have been used to augment synthetic aperture radar (SAR) data for deep learning. However, an automatic target recognition (ATR) model struggles with domain shift when using synthetic data because the model learns specific clutter patterns present in such data, which disturbs performance when applied to measured data with different clutter distributions. This study proposes IRASNet, a domain-generalized SAR-ATR framework designed to achieve effective feature-level clutter reduction and domain-invariant feature learning. The proposed framework introduces a clutter reduction module (CRM) that enhances the signal-to-clutter ratio on feature maps, mitigating the impact of clutter while preserving essential target and shadow information to improve ATR performance. To further enhance generalization, adversarial learning is integrated with CRM to extract clutter-reduced domain-invariant features, bridging the gap between synthetic and measured datasets without requiring measured data during training. Additionally, a positional supervision task is introduced using mask-based ground truth encoding to improve feature extraction from target and shadow regions, strengthening the model's class discrimination capability. Our proposed IRASNet not only enhances generalization performance but also significantly improves feature-level clutter reduction, making it a valuable advancement in the field of radar image pattern recognition.
No takes yet. Share an insight, caveat, or question.
Jang et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: