Research on Sea Surface Ship Detection Algorithm Based on Improved YOLOv8n
Industry Science and Engineering · 2026-04 · 3 (4) · 18-25
DOI: 10.62381/i265403 ↗Kurz gefasst
Fachartikel in der Zeitschrift Industry Science and Engineering, 2026-04. Verfasst von Yuxin Sun. Die Arbeit nennt 21 Literaturangaben.
Abstract (Originalsprache)
Aiming at the problems of complex weather interference, small targets easily obscured by noise, target occlusion, and category imbalance in marine object detection for unmanned surface vehicles (USVs), this paper proposes an improved YOLOv8n-based detection algorithm for marine ships. Taking the lightweight YOLOv8n model as the baseline, the algorithm enhances detection performance in complex marine scenarios through three major strategies: reconstructing an adaptive-weight Focal Loss, embedding the CBAM hybrid-domain attention mechanism, and designing the C2f-DA feature extraction module.Experimental results on a self-built marine object dataset and the Singapore Marine Dataset (SMD) show that, compared with the original YOLOv8n, the improved algorithm increases mAP@0.5 by 2.57% and 1.5% respectively. It exhibits significantly superior detection accuracy and generalization ability for small targets, occluded targets, and complex weather conditions over the baseline model and other mainstream lightweight detection algorithms, while maintaining lightweight characteristics suitable for embedded deployment on unmanned surface vehicles.
Zitieren
APA
Sun, Y. (2026). Research on Sea Surface Ship Detection Algorithm Based on Improved YOLOv8n. Industry Science and Engineering, 3(4), 18-25. https://doi.org/10.62381/i265403
Vancouver
Sun Y. Research on Sea Surface Ship Detection Algorithm Based on Improved YOLOv8n. Industry Science and Engineering. 2026;3(4):18-25. doi:10.62381/i265403
BibTeX
@article{Sun2026,
author = {Sun, Yuxin},
title = {Research on Sea Surface Ship Detection Algorithm Based on Improved YOLOv8n},
journal = {Industry Science and Engineering},
year = {2026},
volume = {3},
number = {4},
pages = {18--25},
doi = {10.62381/i265403}
}
EUVIMED