[1]朱昭云,张 波,叶昭良,等.基于 YOLOv5s 的海上风电设施检测与预警评估[J].计算机技术与发展,2023,33(04):182-189.[doi:10. 3969 / j. issn. 1673-629X. 2023. 04. 027]
　ZHU Zhao-yun,ZHANG Bo,YE Zhao-liang,et al.Offshore Wind Power Facilities Detection and Early Warning Assessment Based on Improved YOLOv5s[J].,2023,33(04):182-189.[doi:10. 3969 / j. issn. 1673-629X. 2023. 04. 027]

点击复制

基于 YOLOv5s 的海上风电设施检测与预警评估()

分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

33

期数:

2023年04期

页码:

182-189

栏目:

人工智能

出版日期:

2023-04-10

文章信息/Info

Title:

Offshore Wind Power Facilities Detection and Early Warning Assessment Based on Improved YOLOv5s

文章编号:

1673-629X(2023)04-0182-08

作者:

朱昭云¹ ; 张 波² ; 叶昭良² ; 黄曙荣¹ ; 曹 卫¹

1. 盐城工学院 机械工程学院,江苏 盐城 224007;

2. 华能海上风电科学技术研究有限公司,江苏 盐城 224045

Author(s):

ZHU Zhao-yun1 ; ZHANG Bo2 ; YE Zhao-liang2 ; HUANG Shu-rong1 ; CAO Wei1

1. School of Mechanical Engineering,Yancheng Institute of Technology,Yancheng 224007,China;

2. Huaneng Offshore Wind Power Science and Technology Research Co. ,Ltd. ,Yancheng 224045,China

关键词:

机器视觉; 风电设施检测; 船舶检测; YOLOv5s; 轻量化; 注意力机制; 双向金字塔

Keywords:

machine vision; detection of wind power facilities; ship detection; YOLOv5s; lightweight; attention mechanism;  bidirectional feature pyramid

分类号:

TP391

DOI:

10. 3969 / j. issn. 1673-629X. 2023. 04. 027

摘要:

海上风电设施检测在海上风电安全监测领域发挥着重要作用。 由于海上环境复杂及海上目标的多样性,现有海上

目标检测算法存在网络复杂、检测精度低的问题,难以满足实时性检测和实用性部署要求。 针对该问题,该文提出

了一种改进 YOLOv5s 的海上风电设施检测算法。 首先,将 YOLOv5s 的主干网络替换为轻量化 GhostNet 进行特征提取,降低网络模型的参数量和计算量;其次,在主干网络末端和 Neck 层分别施加注意力机制( SENet) ,自适应学习重要通道特征权重,提高检测精度;最后,将 Neck 层的 PANet 结构改进为双向金字塔( BiFPN) ,通过融合不同尺度特征提升检测速度。 实验结果表明,该算法在降低网络模型参数量和计算量的同时,在船舶数据集上平均精度达到了 96. 8% ,比原始 YOLOv5s 网络提升了 2. 6 百分点,检测速度达到了 47 FPS。

Abstract:

Offshore wind power facility detection plays a vital role in the field of offshore wind power safety monitoring.?

Due to the complexity of the offshore environment and the diversity of offshore targets, the existing offshore target detection algorithms have theproblems of complex network and low detection accuracy,which are difficult to meet the requirements of real-time detection and practicaldeployment. To solve this problem,we propose an improved YOLOv5s offshore wind power facility detection algorithm. Firstly,thebackbone network of YOLOv5s is replaced by lightweight GhostNet for feature extraction,which reduces the amount of parameters and calculation of the network model. Secondly,the attention mechanism ( SENet) is applied at the end of the backbone network and theNeck layer respectively to adaptively learn the feature weights of important channels to improve the detection accuracy. Finally, thePANet structure of the neck layer is improved to a bi-directional pyramid ( BiFPN) to improve the detection speed by fusing differentscale features. The experimental results show that the proposed algorithm reduces the amount of network model parameters andcalculation,while the average accuracy on the ship data set reaches 96. 8% ,which is 2. 6% higher than the original YOLOv5s network,and the detection speed reaches 47 FPS.

常用功能