Physics-Informed Radar Weak Target Detection via Signal Structural Information-Guided Deep Binary Classification

Weak target detection in low-SNR and cluttered radar environments remains challenging because conventional constant false alarm rate (CFAR) detectors are vulnerable to heterogeneous backgrounds, while purely data-driven deep models often fail to fully exploit radar-domain physical structures. To address this issue, this paper proposes a physics-informed intelligent detection framework for radar weak targets based on signal structural information (SSI)-guided deep binary classification. Specifically, a low-threshold CA-CFAR stage is first employed to generate candidate target coordinates with high recall. An SSI extractor then converts local range–Doppler neighborhoods into structure-preserving slices that retain the characteristic cross-shaped signatures induced by pulse compression and coherent Doppler integration. Based on these SSI slices, the original detection problem is reformulated as a binary classification task. A dedicated classification network is further developed by integrating an anisotropic feature enhancement block, dual-residual perception blocks, channel attention, and multiscale feature aggregation to capture the non-isotropic and scale-varying patterns of target-related structures. Extensive experiments on both simulated and real-measured datasets demonstrate that the proposed framework achieves superior robustness, higher detection probability, and stronger false-alarm suppression than conventional CFAR methods and representative deep-learning baselines. Ablation studies further verify that each architectural component contributes complementary performance gains. The results indicate that coupling radar-domain physical priors with deep representation learning provides an effective paradigm for intelligent weak-target sensing in complex environments.