Multiscale feature and adaptive density-based detector for targets concealed in sea clutter of shipborne HFSWR

High-frequency surface wave radar (HFSWR) exploits high-frequency electromagnetic waves to achieve over-the-horizon maritime target detection and plays a vital role in marine resource monitoring, ocean environment sensing, and maritime security assurance. However, owing to the complex electromagnetic environment over the ocean, its target detection performance is often severely affected by various types of interferences. To address the challenging problem of target detection in sea clutter facing HFSWR, this study proposes a cascaded target detection algorithm. First, considering the complex regional morphology and significant shape variations of sea clutter in the range–doppler (RD) spectrum of the HFSWR, a lightweight multiscale feature fusion semantic segmentation method, termed multiscale lightweight SegNet (MSL-SegNet), is developed. This method effectively reduces the model parameters while maintaining high segmentation accuracy. Subsequently, for the candidate target point set obtained after cell averaging constant false alarm rate (CA-CFAR) detection and extreme learning machine (ELM) filtering, an adaptive parameter and anisotropic (APA)-DBSCAN was proposed. Experimental results based on shipborne HFSWR field data demonstrate that the proposed method achieves effective detection performance.