基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法

doi:10.11930/j.issn.1004-9649.202304059

中国电力 ›› 2023, Vol. 56 ›› Issue (10): 43-52.DOI: 10.11930/j.issn.1004-9649.202304059

• 风电机组及场站主动支撑与运行控制监测关键技术 • 上一篇下一篇

基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法

李冰(), 白云山(), 赵宽(), 郭聪彬(), 翟永杰()

华北电力大学自动化系，河北保定　071003

收稿日期:2023-04-19 出版日期:2023-10-28 发布日期:2023-10-31
作者简介:李冰（1977—），男，副教授，硕士生导师，从事电力视觉研究，E-mail： li_bing_hb@126.com
白云山（1998—），男，硕士研究生，从事模式识别与电力视觉研究，E-mail： 220212216009@ncepu.edu.cn
赵宽（1998—），男，硕士研究生，从事模式识别与电力视觉研究，E-mail： 220212216027@ncepu.edu.cn
郭聪彬（1999—），男，硕士研究生，从事模式识别与电力视觉研究，E-mail： guocongbin@ncepu.edu.cn
翟永杰（1972—），男，通信作者，教授，博士生导师，从事电力视觉研究，E-mail： zhaiyongjie@ncepu.edu.cn
基金资助:
国家自然科学基金联合基金资助项目（可视度受限环境下跨光谱多传感信息融合的机器人语义感知与交互协作，U21A20486）；中央高校基本科研业务费专项资金资助（20237488）。

Surface Defect Detection Algorithm for Wind Turbine Blades Based on HSCA-YOLOv7

Bing LI(), Yunshan BAI(), Kuan ZHAO(), Congbin GUO(), Yongjie ZHAI()

Department of Automation, North China Electric Power University, Baoding 071003, China

Received:2023-04-19 Online:2023-10-28 Published:2023-10-31
Supported by:
This work is supported by National Natural Science Foundation of China Joint Fund Project Highlights (Semantic Perception and Interactive Cooperation of Robots in Limited-Visibility Environments with Cross-Spectral Multi-Sensor Information Fusion, No.U21A20486), the Fundamental Research Funds for the Central Universities (No.20237488).

摘要/Abstract

摘要：

叶片是风电机组的关键部件之一，易受到自然环境因素的影响，出现胶衣脱落、裂纹、腐蚀等损伤，影响风力发电效率及风电机组运行安全。针对航拍风电机组叶片图像缺陷尺度不一、定位不准确、检测精度低等问题，提出了一种HSCA-YOLOv7的风电机组叶片缺陷检测算法。首先根据无人机采集的风电机组叶片图像，制作叶片数据集，采用Mosaic、MixUp方法进行数据扩增；然后将不同膨胀率的深度可分离卷积引入改进空间金字塔池化（improved spatial pyramid pooling，ISPP）模块，减少池化操作带来的细节损失；提出混合空间通道注意力（hybrid spatial channel attention，HSCA）机制，捕获全局视觉场景上下文，增大目标特征与环境语义差异，解决航拍叶片图像缺陷尺度不一的问题；采用Focal EIoU损失函数，解决预测框长宽被错误放大的问题，提高模型对叶片缺陷的定位能力。实验结果表明，所提算法的均值平均精度、均值平均召回率分别达到83.64%、71.96%，与YOLOv7基线算法相比分别提高3.37%、5%。

关键词: 风电机组叶片, 缺陷检测, YOLOv7, 注意力机制, Focal EIoU

Abstract:

The blade is one of the key components of the wind turbine, which is vulnerable to the impact of natural environmental factors, resulting in gel coat falling off, cracks, corrosion, and other damage and thus affecting the efficiency of wind power generation and the safety of wind turbine operation. A defect detection algorithm for wind turbine blades based on HSCA-YOLOv7 is proposed to address the issues of inconsistent defect scale, inaccurate positioning, and low detection accuracy in wind turbine blade images by aerial photography. Firstly, based on the images of wind turbine blades collected by drones, a dataset of blades is created, and Mosaic and MixUp methods are used for data amplification. Then, deep separable convolutions with different expansion rates are introduced into the improved spatial pyramid pooling (ISPP) module to reduce the loss of details caused by pooling operations. Hybrid spatial channel attention (HSCA) is proposed to capture the global visual scene context, increase the semantic difference between target features and the environment, and solve the problem of inconsistent defect scales in blade images. The focal EIoU loss function is used to solve the problem that the length and width of the prediction box are wrongly amplified and improve the positioning ability of the model for blade defects. The experimental results show that the mAP and mAR of the proposed algorithm reach 83.64% and 71.96%, respectively, which are 3.37% and 5% higher than the YOLOv7 baseline algorithm.

Key words: wind turbine blades, defect detection, YOLOv7, attention mechanism, Focal EIoU

李冰, 白云山, 赵宽, 郭聪彬, 翟永杰. 基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法[J]. 中国电力, 2023, 56(10): 43-52.

Bing LI, Yunshan BAI, Kuan ZHAO, Congbin GUO, Yongjie ZHAI. Surface Defect Detection Algorithm for Wind Turbine Blades Based on HSCA-YOLOv7[J]. Electric Power, 2023, 56(10): 43-52.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.electricpower.com.cn/CN/10.11930/j.issn.1004-9649.202304059

https://www.electricpower.com.cn/CN/Y2023/V56/I10/43

图/表 12

图 1 基于HSCA-YOLOv7的风电机组叶片缺陷检测算法

Fig.1 Defect detection algorithm for wind turbine blades based on HSCA-YOLOv7

图 2 ISPP特征加强模块

Fig.2 ISPP feature enhancement module

图 3 深度空洞卷积过程

Fig.3 Depthwise atrous convolution process

图 4 逐点卷积过程

Fig.4 Pointwise convolution process

图 5 HSCA注意力机制

Fig.5 HSCA mechanism

图 6 无人机巡航轨迹与风电机组叶片缺陷类型

Fig.6 Drone cruise track and types of wind turbine blade defects

图 7 Mosaic和MixUp数据增强

Fig.7 Mosaic and MixUp data augmentation

表 1 不同改进策略对评价指标的影响

Table 1 Impact of different improvement strategies on evaluation indicators

组别	改进策略	Macro F1	mAR/%	mAP/%
1	YOLOv7	0.77	66.96	80.27
2	YOLOv7+(1)	0.78	70.05	80.90
3	YOLOv7+(2)	0.78	68.27	81.51
4	YOLOv7+(3)	0.79	70.35	81.41
5	YOLOv7+(1)+(2)	0.79	70.90	81.98
6	YOLOv7+(1)+(3)	0.78	70.25	83.03
7	YOLOv7+(2)+(3)	0.80	71.36	82.33
8	YOLOv7+GAM	0.73	61.24	78.57
9	本文算法	0.81	71.96	83.64

图 8 改进前后算法特征提取热力图视觉对比

Fig.8 Visual comparison of algorithm feature extraction heat map before and after improvement

表 2 不同注意力机制对评价指标的影响

Table 2 Impact of different attention mechanisms on evaluation indicators

采用的注意力机制	Macro F1	mAR/%	mAP/%
NAM	0.69	58.98	76.01
GAM	0.75	71.36	79.69
CBAM	0.77	68.28	79.97
HSCA	0.81	71.96	83.64

表 3 实验结果对比

Table 3 Comparison of experimental results

改进策略	AP50/%			mAP/ %	Param/ M	GFLOPs	FPS/ (f·s^–1)
改进策略	胶衣脱落	裂纹	表面腐蚀	mAP/ %	Param/ M	GFLOPs	FPS/ (f·s^–1)
YOLOv5	76.10	71.90	84.00	77.33	7.02	15.8	166.7
YOLOv7	81.43	75.09	84.29	80.27	37.21	52.6	77.4
YOLOv8	81.80	63.00	77.80	74.20	3.00	8.1	192.3
Faster RCNN	81.89	67.69	79.53	76.37	41.37	164.9	12.8
本文算法	90.60	78.90	81.43	83.64	34.86	53.5	65.0

图 9 不同算法结果对比

Fig.9 Comparison of results of different algorithms

参考文献 25

1	时智勇, 王彩霞, 李琼慧. “十四五”中国海上风电发展关键问题[J]. 中国电力, 2020, 53 (7): 8- 17.
	SHI Zhiyong, WANG Caixia, LI Qionghui. Key issues of China’s offshore wind power development in the “14 th five-year plan”[J]. Electric Power, 2020, 53 (7): 8- 17.
2	胡蓓, 吴永康, 郭子君, 等. 风力发电叶片裂缝监测技术综述[J]. 高压电器, 2022, 58 (7): 93- 100.
	HU Bei, WU Yongkang, GUO Zijun, et al. Review of wind turbine blade crack monitoring technology[J]. High Voltage Apparatus, 2022, 58 (7): 93- 100.
3	海涛, 范恒, 王楷杰, 等. 基于PSO-SVM算法的风电机组结冰故障诊断[J]. 智慧电力, 2021, 49 (4): 1- 6, 74.
	HAI Tao, FAN Heng, WANG Kaijie, et al. Fault diagnosis of wind turbine icing based on PSO-SVM algorithm[J]. Smart Power, 2021, 49 (4): 1- 6, 74.
4	WANG W J, XUE Y, HE C K, et al. Review of the typical damage and damage-detection methods of large wind turbine blades[J]. Energies, 2022, 15 (15): 5672. DOI
5	MCGUGAN M, MISHNAEVSKY L. Damage mechanism based approach to the structural health monitoring of wind turbine blades[J]. Coatings, 2020, 10 (12): 1223. DOI
6	黄子恒, 许钊源, 伍剑波, 等. 基于优化模态分解和Xgblr的风机叶片故障诊断方法[J]. 机械设计, 2022, 39 (7): 56- 62.
	HUANG Ziheng, XU Zhaoyuan, WU Jianbo, et al. Fault diagnosis method of fan blade based on optimized modal decomposition and Xgblr[J]. Journal of Machine Design, 2022, 39 (7): 56- 62.
7	AWADALLAH M, EL-SINAWI A. Effect and detection of cracks on small wind turbine blade vibration using special Kriging analysis of spectral shifts[J]. Measurement, 2020, 151, 107076. DOI
8	葛畅, 阎洁, 刘永前, 等. 海上风电场运行控制维护关键技术综述[J]. 中国电机工程学报, 2022, 42 (12): 4278- 4292.
	GE Chang, YAN Jie, LIU Yongqian, et al. Review of key technologies for operation control and maintenance of offshore wind farm[J]. Proceedings of the CSEE, 2022, 42 (12): 4278- 4292.
9	陶显, 侯伟, 徐德. 基于深度学习的表面缺陷检测方法综述[J]. 自动化学报, 2021, 47 (5): 1017- 1034.
	TAO Xian, HOU Wei, XU De. A survey of surface defect detection methods based on deep learning[J]. Acta Automatica Sinica, 2021, 47 (5): 1017- 1034.
10	孙月莹, 陈俊霖, 张胜茂, 等. 基于改进YOLOv7的毛虾捕捞渔船作业目标检测与计数方法[J]. 农业工程学报, 2023, 39 (10): 151- 162. DOI
	SUN Yueying, CHEN Junlin, ZHANG Shengmao, et al. Target detection and counting method for Acetes chinensis fishing vessels operation based on improved YOLOv7[J]. Transactions of the Chinese Society of Agricultural Engineering, 2023, 39 (10): 151- 162. DOI
11	张艳君, 沈平, 郭安辉, 等. 融合CBAM-YOLOv7模型的路面缺陷智能检测方法研究[J/OL]. 重庆理工大学学报(自然科学): 1–7[2023-07-04]. http://kns.cnki.net/kcms/detail/50.1205.T.20230629.1123.002.html.
	ZHANG Yanjun, SHEN Ping, GUO Anhui, et al. . Research on intelligent detection method of pavement defects incorporating CBAM-YOLOv7 model[J/OL]. Journal of Chongqing University of Technology (Natural Science): 1–7[2023-07-04]. http://kns.cnki.net/kcms/detail/50.1205.T.20230629.1123.002.html.
12	仝卫国, 仪小龙, 李冰, 等. 融合多尺度特征与注意力机制的风机桨叶缺陷检测方法[J]. 电子测量技术, 2022, 45 (24): 166- 172.
	TONG Weiguo, YI Xiaolong, LI Bing, et al. Fan blade defect detection method combining multi-scale features and attention mechanism[J]. Electronic Measurement Technology, 2022, 45 (24): 166- 172.
13	GUO J H, LIU C, CAO J F, et al. Damage identification of wind turbine blades with deep convolutional neural networks[J]. Renewable Energy, 2021, 174, 122- 133. DOI
14	ZHANG J J, COSMA G, WATKINS J. Image enhanced mask R-CNN: a deep learning pipeline with new evaluation measures for wind turbine blade defect detection and classification[J]. Journal of Imaging, 2021, 7 (3): 46. DOI
15	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[EB/OL](2022-07-15)[2022-12-21].https://arxiv.org/abs/2207.02696.
16	GE Z, LIU S, WANG F, et al. Yolox: Exceeding yolo series in 2021[J/OL]. 2021. DOI:10.48550/arXiv.2107.08430.
17	郑果, 姜玉松, 沈永林. 基于改进YOLOv7的水稻害虫识别方法[J]. 华中农业大学学报, 2023, 42 (3): 143- 151.
	ZHENG Guo, JIANG Yusong, SHEN Yonglin. Recognition of rice pests based on improved YOLOv7[J]. Journal of Huazhong Agricultural University, 2023, 42 (3): 143- 151.
18	LIU Y C, SHAO Z R, HOFFMANN N. Global attention mechanism: retain information to enhance channel-spatial interactions[EB/OL](2021-12-10)[2022-12-21].https://arxiv.org/abs/2112.05561.
19	ZHANG Y F, REN W Q, ZHANG Z, et al. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506, 146- 157. DOI
20	谭兴国, 张高明. 基于无人机巡检的风机叶片表面缺陷检测技术[J/OL]. 电测与仪表: 1–10[2023-04-10]. http://kns.cnki.net/kcms/detail/23.1202.T H.20221013.1050.004.html.
	TAN Guoxing, ZHANG Gaoming. UAV-based inspection of wind turbine blade surface defects detection technology[J/OL]. Electrical Measure-ment & Instrumentation: 1–10[2023-02-02]. http:// kns.cnki.net/ kcms/de-tail/23.1202.TH.20221013. 1050.004.html.
21	焦嵩鸣, 白健鹏, 首云锋. 风机叶片精准巡视的无人机控制策略研究[J]. 中国电机工程学报, 2023, 43 (10): 3822- 3832.
	JIAO Songming, BAI Jianpeng, SHOU Yunfeng. Research on UAV control strategy for accurate inspection of wind turbine blades[J]. Proceedings of the CSEE, 2023, 43 (10): 3822- 3832.
22	赵文清, 杨盼盼. 双向特征融合与注意力机制结合的目标检测[J]. 智能系统学报, 2021, 16 (6): 1098- 1105.
	ZHAO Wenqing, YANG Panpan. Target detection based on bidirectional feature fusion and an attention mechanism[J]. CAAI Transactions on Intelligent Systems, 2021, 16 (6): 1098- 1105.
23	WANG J W, XU C, YANG W, et al. A normalized Gaussian Wasserstein distance for tiny object detection[EB/OL](2021-10-26)[2022-12-21].https://arxiv.org/abs/2110.13389.
24	LIU Y C, SHAO Z R, TENG Y Y, et al. NAM: normalization-based attention module[EB/OL](2021-11-24)[2022-12-21].https://arxiv.org/abs/2111.12419.
25	黄冬梅, 王玥琦, 胡安铎, 等. 融合多维度特征的绝缘子状态边缘识别方法[J]. 中国电力, 2022, 55 (1): 133- 141.
	HUANG Dongmei, WANG Yueqi, HU Anduo, et al. An edge recognition method for insulator state based on multi-dimension feature fusion[J]. Electric Power, 2022, 55 (1): 133- 141.

[1]	宋智伟, 黄新波, 纪超, 张凡, 张烨. 基于PCSA-YOLOv7 Former的输电线路连接金具及其锈蚀检测方法[J]. 中国电力, 2024, 57(6): 141-152.
[2]	高岩, 吴汉斌, 张纪欣, 张华铭, 张沛. 基于组合深度学习的光伏功率日前概率预测模型[J]. 中国电力, 2024, 57(4): 100-110.
[3]	吴晓刚, 阎洁, 葛畅, 唐雅洁, 倪筹帷, 季青锋. 基于改进GRU-CNN的风光水一体化超短期功率预测方法[J]. 中国电力, 2023, 56(9): 178-186,205.
[4]	张鑫, 叶俊杰, 崔瑶, 黄鑫, 仲林林. 基于语义信息距离解耦的变电运维多类别缺陷图像检测[J]. 中国电力, 2023, 56(6): 209-218.
[5]	陆友文, 崔昊, 陈佳宁, 彭祥佳, 冯双, 刘栋. 基于RA-CNN和同步相量的风电场次/超同步振荡参数智能辨识方法[J]. 中国电力, 2023, 56(4): 46-55,67.
[6]	周震震, 宋云海, 何宇浩, 王黎伟, 黄和燕, 何珏, 朱志航, 闫云凤. 基于分组查询注意力的可扩展电力人员行为分类方法[J]. 中国电力, 2023, 56(11): 77-85.
[7]	苏向敬, 宇海波, 符杨, 田书欣, 李海瑜, 耿福海. 基于DALSTM和联合分位数损失的海上风电功率概率预测[J]. 中国电力, 2023, 56(11): 10-19.
[8]	陈铁, 张治藩, 李咸善, 陈一夫, 李鸿鑫. 基于混合模态分解和LSTM-CNN的变压器油中溶解气体浓度预测[J]. 中国电力, 2023, 56(1): 132-141.
[9]	樊江川, 于昊正, 刘慧婷, 杨丽君, 安佳坤. 基于多分支门控残差卷积神经网络的短期电力负荷预测[J]. 中国电力, 2022, 55(11): 155-162,174.
[10]	刘逸凡, 王淑青, 庆毅辉, 王晨曦, 兰天泽, 要若天. 基于EfficientDet和双目摄像头的绝缘子缺陷检测[J]. 中国电力, 2021, 54(2): 156-163,196.
[11]	黄锐勇, 戴美胜, 郑跃斌, 黄勤琴, 康立烨, 苟先太, 周维超. 电力设备红外图像缺陷检测[J]. 中国电力, 2021, 54(2): 147-155.
[12]	曹培, 徐鹏, 贺建明, 高凯, 田昊洋, 季怡萍. 基于多源感知的开关柜绝缘缺陷检测技术[J]. 中国电力, 2021, 54(10): 117-124,133.
[13]	代真, 李伟, 郝晓军, 刘红权, 敬尚前. 发电机护环内壁缺陷TOFD检测优化[J]. 中国电力, 2014, 47(3): 65-68.

基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法

Surface Defect Detection Algorithm for Wind Turbine Blades Based on HSCA-YOLOv7

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 13

编辑推荐

Metrics

模态框（Modal）标题

基于HSCA-YOLOv7的风电机组叶片表面缺陷检测算法

Surface Defect Detection Algorithm for Wind Turbine Blades Based on HSCA-YOLOv7

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 13

编辑推荐

Metrics