Answers are grounded in this paper and cite their sources.
In ship detection using high-frequency surface wave radar, traditional constant false alarm rate (CFAR) detectors rely on manual thresholding, which leads to significant performance degradation in clutter regions and under strong interference. This article proposes a detection framework that delivers both high accuracy and real-time processing by operating directly on Range–Doppler (RD) spectra. The core of our solution is HFR-YOLO, a novel model built upon the YOLOv10 architecture. It incorporates several key modules to enhance the ability to discriminate weak targets from clutter, capture multiscale targets, and ensure robustness across diverse scenarios. These improvements facilitate continuous track extraction from radar data. To overcome the scarcity of real training samples within the Bragg peak region, we incorporate a simulation-measurement integrated transfer learning strategy, realized through a two-stage progressive training scheme for the HFR-YOLO model, to enhance its generalization in realistic cluttered environments. Experimental results on measured radar data show that the HFR-YOLO achieves higher precision and recall than other YOLO variants and several baseline models; moreover, the proposed detection framework improves the detection rate by 22.16% and reduces the false-alarm rate by 31.43% compared to CFAR detectors.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: