火力与指挥控制2026，Vol.51Issue(3)：44-49,58,7.DOI:10.3969/j.issn.1002-0640.2026.03.006
基于优化YOLOv8的SAR图像舰船目标检测算法
An Improved YOLOv8-based Algorithm for Ship Target Detection in SAR Images
摘要
Abstract
To address the problems of noise interference and poor performance in detecting small tar-gets in ship target detection in SAR images,the AD-YOLO algorithm is proposed.Based on YOLOv8,the algorithm introduces the ADNet for denoising before the target detection network to improve image qual-ity,replaces the C2f module in the backbone network with the DWR_C2f module to enhance the feature representation of multi-scale targets,and introduces the DAttention attention mechanism after the SPPF to adapt to complex scenarios.Experiments on the SARDet-100k dataset show that compared with the baseline model YOLOv8,AD-YOLO achieves improvements of 1.33 and 1.00 in mAP in the n and s model sizes respectively.Moreover,it exhibits stronger robustness against background noise,small tar-gets,scattering interference and other situations,thereby effectively enhancing the detection accuracy and robustness of ship targets in SAR images.关键词
SAR图像/目标检测/Yolov8/ADNet/DAttention/DWRKey words
SAR images/target detection/YOLOv8/ADNet/DAttention/DWR分类
信息技术与安全科学引用本文复制引用
郑志材..基于优化YOLOv8的SAR图像舰船目标检测算法[J].火力与指挥控制,2026,51(3):44-49,58,7.基金项目
广东省普通高校重点领域专项(2024ZDZX4156) （2024ZDZX4156）
广东省普通高校特色创新类基金资助项目(2024KTSCX202) （2024KTSCX202）
