摘要
受测量机制和海洋环境因素影响,侧扫声纳沉船目标存在多尺度、多畸变、分辨率低的问题,严重影响了SSS沉船目标检测精度。为此,提出一种基于改进YOLOv8的侧扫声纳沉船目标检测方法。首先,为了优化网络的特征提取能力,在YOLOv8主干网络中引入CMUNeXt模块,增强了网络对不同尺度、形状和背景噪声的侧扫声纳目标的适应能力。之后,在网络中引入多尺度特征融合模块,并行处理不同尺度的沉船目标特征,并采用加权融合的机制实现差异化特征融合,增强网络的特征表达能力。实验表明,本文所提出方法在各项定量和定性评价中均超越了同系列的YOLO网络。相较于YOLOv8网络,精确率、召回率和mAP分别提升了5.0%,13.0%和7.8%。
Due to the influence of measurement mechanisms and marine environmental factors,side-scan sonar(SSS)shipwreck targets have problems such as multi-scale,multi-distortion,and low resolution,which seriously affect the detection accuracy of SsS shipwreck targets.Therefore,this paper proposes a side-scan sonar shipwreck target detection method based on the improved YOLOv8.Firstly,to optimize the feature extraction ability of the network,the CMUNeXt module is introduced into the backbone network of YOLOv8,enhancing the network's adaptability to side-scan sonar targets of different scales,shapes,and background noise.Then,a multi-scale feature fusion module is introduced into the network to process the features of shipwreck targets of different scales in parallel,and a weighted fusion mechanism is adopted to achieve differentiated feature fusion,enhancing the network's feature expression ability.Experiments show that the proposed method outperforms the YOLO network of the same series in all quantitative and qualitative evaluations.Compared with the YOLOv8 network,the precision,recall rate,and mAP have increased by 5.0%,13.0%,and 7.8%respectively.
作者
郭岳山
郭爽
GUO Yueshan;GUO Shuang(DEEPINFAR,Inc.,Tianjin 300459,China;Wuhan Changjiang waterway Rescue and Salvage Bureau,Wuhan 430010,China)
出处
《海洋测绘》
北大核心
2025年第3期31-35,共5页
Hydrographic Surveying and Charting
