SSWR-YOLO: Lightweight Vessel Detection for Shipborne HFSWR in Complex Maritime Scenarios
Shipborne high-frequency surface wave radar (SHFSWR) plays a critical role in maritime surveillance but suffers from severe performance degradation in complex maritime scenarios characterized by sea clutter, dense target aggregation, and weak echo returns. To address these challenges, we propose Shipborne Surface Wave Radar You Only Look Once (SSWR-YOLO), a lightweight detection network built upon the YOLOv8 framework with specialized modules for robust detection under adverse conditions. Specifically, a receptive field spatial attention convolution is embedded into the backbone to dynamically enhance target responses relative to broadened sea clutter and highlight weak target features. Furthermore, a context aggregation network is integrated into the neck to enhance multi-scale feature fusion, effectively resolving target masking in dense vessel target scenarios. Additionally, a softened non-maximum suppression strategy is employed to prevent the erroneous removal of overlapped vessel proposals. Experimental results on both simulated and field SHFSWR data demonstrate that, on the 50 held-out test images, SSWR-YOLO achieves a 9.76 percentage point improvement in detection rate and a 3.11% reduction in overall false discovery rate over the baseline YOLOv8, while introducing only marginal parameter overhead. The proposed method offers an efficient and reliable solution for vessel detection under challenging maritime conditions, with a core network forward pass compatible with near-real-time processing on a desktop GPU and potential for embedded deployment.
Authors
- Yiming Wang (ORCID: https://orcid.org/0000-0002-0935-3983)
- Mengni Zhang (ORCID: https://orcid.org/0000-0001-7437-5209)
- Henglin Wang
- Xifang Jin
- Chengzhen Ji
- Yingying Zhang
- Haoran Xu (ORCID: https://orcid.org/0009-0002-1693-6017)
Institutions
- Qilu University of Technology (CN)
- Harbin Institute of Technology (CN)
- Ministry of Natural Resources (CN)
- First Institute of Oceanography (CN)
- Institute of Oceanographic Instrumentation (CN)
- Shandong Academy of Sciences (CN)
Publication Details
- Journal
- Remote Sensing
- Published
- 2026-11-20
- DOI
- https://doi.org/10.3390/rs18223998
- Primary Topic
- Advanced Neural Network Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China