设为首页 联系我们 加入收藏

当前位置: 网站首页 期刊分类目录 2025年第2期 正文

基于改进YOLO v5s的水下小目标检测算法

作者:时间:2025-05-02点击数:



全文下载: 17.pdf

文章编号: 1672-6987202502-0122-10 DOI 10.16351/j.1672-6987.2025.02.017


梁衡1a2 刘儒一1a  1b 宋廷强1a*1. 青岛科技大学  a. 信息科学技术学院; b. 自动化与电子工程学院, 山东 青岛 2660612. 烟台港股份有限公司, 山东 烟台 264001

摘要: 针对目前水下图像存在图像模糊以及小目标聚集导致水下小目标识别精度低的情况提出一种基于改进YOLO v5s的水下小目标检测算法在主干特征提取网络中嵌入卷积注意力模块强化小目标信息提高网络模型的特征提取能力设计了一种改进的C3模块C3Swin在原始C3模块中加入Swin Transformer结构在不同滑动窗口间进行信息交互增强了全局信息的提取能力对原始YOLO v5s的检测层进行重构增加小目标检测层提升小目标的检测精度改进损失函数使用-iou对原损失函数进行优化提升预测框的回归精度实验结果表明URPC水下目标检测数据集中本工作提出的算法平均精度均值mAP86.9%相较于原模型提升了2.9%检测速度为62.7 Hz优于主流算法在保证检测速度的同时提升了检测精度


关键词: 水下小目标检测 YOLO v5s 卷积注意力模块 Swin Transformer -iou


中图分类号: TP 391        文献标志码: A

引用格式: 梁衡 刘儒一 张典 . 基于改进YOLO v5s的水下小目标检测算法J. 青岛科技大学学报自然科学版 2025 462 122-131.


LIANG Heng LIU Ruyi ZHANG Dian et al. Underwater small object detection algorithm based on improved YOLO v5sJ. Journal of Qingdao University of Science and TechnologyNatural Science Edition 2025 462 122-131.

Underwater Small Object Detection Algorithm Based on Improved YOLO v5s

LIANG Heng1a2 LIU Ruyi1a ZHANG Dian1b SONG Tingqiang1a1. a.College of Information Science and Technologyb. College of Automation and Electronic EngineeringQingdao University of Science and TechnologyQingdao 266061China2. Yantai Port Co.Ltd.Yantai 264001China

Abstract In order to solve the problem of low accuracy of underwater small object recognition caused by underwater image blur and small object aggregationAn underwater small object detection algorithm based on improved YOLO v5s is proposed. The convolution attention module is embedded in the backbone feature extraction network to strengthen the small object information and improve the feature extraction ability of the network model. An improved C3 moduleC3Swinis designed. The Swin Transformer structure is added to the original C3 moduleand information is interacted between different sliding windowswhich enhances the ability to extract global information. The detection layer of the original YOLO v5s was reconstructed. Adding a small object detection layer and improve the detection accuracy of small object. Using α- Iou to optimize the original loss function and improve the regression accuracy of the bounding box. The results show that in the URPC underwater object detection datasetthe average precision mAP of the algorithm proposed in this paper is 86.9%which is 2.9% higher than the original modeland the detection speed is 62.7 Hzwhich is superior to the mainstream algorithm.The proposed algorithm ensures the detection speed and improve the detection accuracy.


Key words underwater object detectionYOLO v5sconvolutional attention moduleSwin Transformer-iou

收稿日期: 2024-12-29

基金项目: 山东省重点研发计划项目2019GGX101047;山东省自然科学基金项目ZR2021MF023.

作者简介: 梁衡1997—,男,硕士研究生.     * 通信联系人.


Copyright © 2011-2017 青岛科技大学学报 (自然科学版)