电子测量与仪器学报

张银胜,杨宇龙,吉茹,蓝天鹤,单慧琳.改进 YOLOv5s 的风力涡轮机表面缺陷检测[J].电子测量与仪器学报,2023,37(1):40-49

改进 YOLOv5s 的风力涡轮机表面缺陷检测

Surface defect detection of wind turbine based on YOLOv5s

DOI：10.13382/j.issn.1000-7105.2023.01.005

中文关键词: 风力涡轮机 YOLOv5s 轻量化目标检测注意力机制多尺度融合

英文关键词:wind turbines YOLOv5s lightweight object detection mechanism of attention multi-scale fusion

基金项目:国家自然科学基金 (62071240,62106111)、江苏省一流本科课程《电路分析基础》无锡学教学改革重点课题( JGZD202109) 项目资助

作者	单位
张银胜	1. 南京信息工程大学电子与信息工程学院，2. 无锡学院电子信息工程学院
杨宇龙	1. 南京信息工程大学电子与信息工程学院
吉茹	1. 南京信息工程大学电子与信息工程学院
蓝天鹤	2. 无锡学院电子信息工程学院
单慧琳	1. 南京信息工程大学电子与信息工程学院，2. 无锡学院电子信息工程学院

Author	Institution
Zhang Yinsheng	1. School of Electronic and Information Engineering, Nanjing University of Information Science and Technology,2. School of Electronic and Information Engineering,Wuxi University
Yang Yulong	1. School of Electronic and Information Engineering, Nanjing University of Information Science and Technology
Ji Ru	1. School of Electronic and Information Engineering, Nanjing University of Information Science and Technology
Lan Tianhe	2. School of Electronic and Information Engineering,Wuxi University
Shan Huilin	1. School of Electronic and Information Engineering, Nanjing University of Information Science and Technology,2. School of Electronic and Information Engineering,Wuxi University

摘要点击次数: 1152

全文下载次数: 1322

中文摘要:

针对传统方式检测风力涡轮机表面缺陷时出现的精度不足、泛化性较差问题,提出了一种改进 YOLOv5s 的风力涡轮机表面缺陷检测模型。在网络结构方面,首先在主干特征提取网络引入改进的 MobileNetv3 网络,用于协调并平衡模型的轻量化和精度关系;其次采用 BiFPN 式的融合方式,增强神经网络的多尺度适应能力,提高融合速度和效率;最后为轻量化的自适应调节特征权重,运用 ECAnet 通道注意力机制,进一步提高神经网络的特征提取能力。在损失函数方面,将边框回归的损失函数修改为 αIoU Loss,进一步提升了 bbox 回归精度。实验结果表明,基于 YOLOv5s 的改进算法可以在复杂环境下快速准确地识别风机表面的缺陷目标,能够满足实时目标检测的实际应用需求。

英文摘要:

Aiming at the problem of insufficient precision and poor generalization in the traditional way of wind turbine surface defect detection, an improved YOLOv5s wind turbine surface defect detection model is proposed. In terms of network structure, an improved MobileNetv3 network is introduced into the backbone feature extraction network to coordinate and balance the lightweight and accuracy relationship of the model. Secondly, the BiFPN fusion method is adopted to enhance the multi-scale adaptability of the neural network and improve the fusion speed and efficiency. Finally, for the lightweight adaptive adjustment of feature weights, the ECAnet channel attention mechanism is used to further improve the feature extraction ability of the neural network. In terms of loss function, the loss function of bounding box regression is modified to αIoU Loss, which further improves the accuracy of bbox regression. The experimental results show that the improved algorithm based on YOLOv5s can quickly and accurately identify the defect targets on the surface of the wind turbine in complex environments, and can meet the practical application requirements of real-time target detection.

查看全文查看/发表评论下载PDF阅读器