基于DBBR-YOLO的光伏电池表面缺陷检测

doi:10.11996/JG.j.2095-302X.2024050913

图学学报 ›› 2024, Vol. 45 ›› Issue (5): 913-921.DOI: 10.11996/JG.j.2095-302X.2024050913

• 图像处理与计算机视觉 • 上一篇下一篇

基于DBBR-YOLO的光伏电池表面缺陷检测

刘义艳¹(), 郝婷楠¹, 贺晨², 常英杰¹()

1.长安大学能源与电气工程学院，陕西西安 710018
2.西安市轨道交通集团有限公司运营分公司，陕西西安 710016

收稿日期:2024-04-09 修回日期:2024-08-14 出版日期:2024-10-31 发布日期:2024-10-31
通讯作者:常英杰(1992-)，男，讲师，博士。主要研究方向为大数据处理、机器学习和多相流等。E-mail：yj.chang@chd.edu.cn
第一作者:刘义艳(1981-)，女，副教授，博士。主要研究方向为电网大数据处理和电能质量分析等。E-mail：yyliu1@chd.edu.cn
基金资助:
陕西省重点研发计划项目(2021GY-098)

Photovoltaic cell surface defect detection based on DBBR-YOLO

LIU Yiyan¹(), HAO Tingnan¹, HE Chen², CHANG Yingjie¹()

1. School of Energy and Electrical Engineering, Chang’an University, Xi’an Shaanxi 710018, China
2. Operation Branch of Xi’an Rail Transit Group Co., Ltd, Xi’an Shaanxi 710016, China

Received:2024-04-09 Revised:2024-08-14 Published:2024-10-31 Online:2024-10-31
Contact: CHANG Yingjie (1992-), lecturer, Ph.D. His main research interests cover processing big data, machine learning and multiphase flow. E-mail：yj.chang@chd.edu.cn
First author：LIU Yiyan (1981-), associate professor, Ph.D. Her main research interests cover processing big data in power grids and analyzing power quality. E-mail：yyliu1@chd.edu.cn
Supported by:
Key Research and Development Program of Shaanxi Province(2021GY-098)

摘要/Abstract

摘要：

针对光伏电池表面缺陷特征提取困难以及检测的实时性和准确性问题，提出了一种基于DBBR-YOLO的光伏电池表面缺陷检测方法。首先，将多样化分支块(DBB)融入到YOLOv8n中Backbone部分的C2f模块中，引入多样化的特征提取路径，增强特征提取的能力；其次，将模型的Neck部分和Gold-YOLO进行融合，实现对不同层级特征的全局信息聚合和融合，提高了特征图之间的信息交互效率，增强了模型的特征表达能力；最后，引入SimAM注意力机制提高了特征的表达能力，以增强模型对微小缺陷或小目标的检测能力。实验选取5种光伏电池表面缺陷类型进行验证，结果表明：改进后的DBBR-YOLO模型mAP50值达到93.1%，相较于YOLOv8n提升了3.7%，FPS值达到了158.0，该模型在精度和速度方面的性能可以满足实时性、准确性的要求，能够应对光伏电池表面缺陷检测的实际应用场景。

关键词: 光伏电池缺陷, Gold-YOLO, 注意力机制, 深度学习, YOLOv8

Abstract:

A method for detecting surface defects of photovoltaic (PV) cells based on DBBR-YOLO was proposed to address the difficulties in defect feature extraction and the issues of real-time detection and accuracy. Firstly, a diverse branch block (DBB) was incorporated into the C2f module of the YOLOv8n Backbone section to introduce diversified feature extraction paths, enhancing the capability of feature extraction. Secondly, the Neck section of the model was fused with Gold-YOLO to achieve global information aggregation and feature fusion at different hierarchical levels, improving the efficiency of information interaction between feature maps and enhancing the feature expression capability of the model. Finally, the SimAM attention mechanism was introduced to improve the feature expression capability, thereby enhancing the model’s ability to detect small defects or targets. Experiments conducted on five types of PV cell surface defects demonstrated that the improved DBBR-YOLO model achieved an mAP50 value of 93.1%, a 3.7% improvement over YOLOv8n, with an FPS value of 158.0. The performance of the model in terms of accuracy and speed can meet the requirements for real-time detection and accuracy, making it suitable for practical application scenarios of detecting PV cell surface defects.

Key words: photovoltaic cell defects, Gold-YOLO, attention mechanism, deep learning, YOLOv8

中图分类号:

TP391

刘义艳, 郝婷楠, 贺晨, 常英杰. 基于DBBR-YOLO的光伏电池表面缺陷检测[J]. 图学学报, 2024, 45(5): 913-921.

LIU Yiyan, HAO Tingnan, HE Chen, CHANG Yingjie. Photovoltaic cell surface defect detection based on DBBR-YOLO[J]. Journal of Graphics, 2024, 45(5): 913-921.

图/表 14

图1 YOLOv8网络结构图

Fig. 1 YOLOv8 network architecture diagram

图2 DBBR-YOLO结构图

Fig. 2 DBBR-YOLO network architecture diagram

图3 Diverse branch block (DBB)结构图

Fig. 3 Diverse branch block (DBB) structure diagram

图4 C2f_DBB结构图

Fig. 4 C2f_DBB structure diagra

图5 低阶收集分支结构

Fig. 5 Low-stage gather-and-distribute branch

图6 低阶分发分支结构

Fig. 6 Low-stage circulate-and-distribute branch

表1 实验环境配置

Table 1 Experimental environment configuration

配置环境	版本
操作系统	Windows 11
CPU	Intel(R) Core(TM) i7-13700KF
显卡	NVIDIA GeForce RTX 4090 24 GB
深度学习框架	Pytorch
CUDA	CUDA 11.2
Python	Python-3.8.18

表2 实验参数设置

Table 2 Experimental parameter settings

超参数	参数值	超参数	参数值
Images size	640	Momentum	0.937
Batch size	32	Epochs	100
Learning rate	0.01	Optimizer	SGD

图7 光伏电池缺陷数据集类型((a)裂纹；(b)黑芯；(c)断栅；(d)粗线；(e)短路)

Fig. 7 Photovoltaic cell defect dataset type ((a) Crack; (b) Black_core; (c) Finger; (d) Thick_line; (e) Short_circuit)

表3 消融实验

Table 3 Melting experiment

Group	Model	参数量/M	mAP50/%	mAP50:95/%	GFLOPs/G	FPS	F1
1	YOLOv8n	3.1	89.4	66.2	8.1	166.2	0.84
2	+C2f_DBB	3.1	90.6	65.3	8.1	173.5	0.86
3	+RepGDNeck	5.9	91.1	66.8	10.2	108.2	0.88
4	+C2f_DBB+RepGDNeck	5.9	92.3	67.7	10.2	90.4	0.88
5	+C2f_DBB+RepGDNeck+simAM	5.9	93.1	68.2	10.2	158.0	0.89

表4 注意力机制对比结果

Table 4 Attention mechanism comparison results

Model	参数量/M	mAP50/%	mAP50:95/%	GFLOPs/G	FPS	F1
YOLOv8n	3.1	89.4	66.2	8.1	166.2	0.84
+DAattent--ion	3.3	91.9	68.5	8.3	91.0	0.88
+MLCA	3.2	90.4	67.9	8.1	99.9	0.86
+CPCA	4.0	91.7	68.2	8.3	150.5	0.88
+simAM	3.1	91.9	69.0	8.3	159.8	0.88

表5 YOLO算法对比实验

Table 5 YOLO algorithm comparative experiment

Model	参数量/M	mAP50/%	mAP50:95/%	GFLOPs/G	FPS	F1
YOLOv3n	103.0	88.5	65.0	282.2	52.0	0.82
YOLOv5n	2.5	89.0	66.0	7.1	168.9	0.83
YOLOv8m	25.0	89.2	68.0	78.7	108.8	0.84
YOLOv8s	11.1	89.4	67.5	28.4	151.0	0.84
YOLOv8n	3.1	89.4	66.2	8.1	166.2	0.84
DBBR-YOLO (Ours)	5.9	93.1	68.2	10.2	158.0	0.89

图8 检测对比结果((a) YOLOv8；(b)本文算法)

Fig. 8 Detection comparison results ((a) YOLOv8; (b) Ours)

图9 漏检、误检、正确的检测效果((a) YOLOv8；(b)本文算法)

Fig. 9 Missed detections, false alarms, and correct detections ((a) YOLOv8; (b) Ours)

参考文献 22

[1]	张治, 邹鹏辉, 刘志锋, 等. 发射极掺杂工艺对产业化IBC太阳电池性能的影响[J]. 太阳能学报, 2022, 43(3): 158-162. DOI
	ZHANG Z, ZOU P H, LIU Z F, et al. Influence of emitter doping process on performance of industrial IBC solar cell[J]. Acta Energiae Solaris Sinica, 2022, 43(3): 158-162 (in Chinese). DOI
[2]	白浩良, 王晨, 卢静, 等. 聚光光伏系统太阳能电池散热技术及发展现状[J]. 化工进展, 2023, 42(1): 159-177. DOI
	BAI H L, WANG C, LU J, et al. Solar cell heat dissipation technology and development status of concentrating photovoltaic system[J]. Chemical Industry and Engineering Progress, 2023, 42(1): 159-177 (in Chinese). DOI
[3]	彭自然, 张颖清, 肖伸平. 基于YOLOv5的太阳电池表面缺陷检测[J]. 太阳能学报, 2024, 45(6): 368-375.
	PENG Z R, ZHANG Y Q, XIAO S P. Research on surface defect detection of solar cell with improved YOLOv5 algorithm[J]. Acta Energiae Solaris Sinica, 2024, 45(6): 368-375 (in Chinese).
[4]	RAJPUT A S, HO J W, ZHANG Y, et al. Quantitative estimation of electrical performance parameters of individual solar cells in silicon photovoltaic modules using electroluminescence imaging[J]. Solar Energy, 2018, 173: 201-208.
[5]	MATSUI T, TATSUMI K, KAWASHIMA T, et al. Sensitivity improvement of ultrasonic beam induced resistance change (SOBIRCH) method with ultrasound resonance inside mold resin[J]. Microelectronics Reliability, 2020, 109: 113641.
[6]	NETZELMANN U, WALLE G, LUGIN S, et al. Induction thermography: principle, applications and first steps towards standardisation[J]. Quantitative InfraRed Thermography Journal, 2016, 13(2): 170-181.
[7]	杜博伦, 何赟泽, 杨瑞珍, 等. 电磁感应对硅光伏电池可视化检测技术的改性[J]. 仪器仪表学报, 2018, 39(10): 158-165.
	DU B L, HE Y Z, YANG R Z, et al. Modification of visual detection for silicon photovoltaic cells based on electromagnetic induction[J]. Chinese Journal of Scientific Instrument, 2018, 39(10): 158-165 (in Chinese).
[8]	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149. DOI PMID
[9]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE Press, 2016: 779-788.
[10]	雷斯达, 林杰, 黄思璐, 等. 基于改进YOLOv3网络对混凝土表面裂缝目标识别研究[J]. 公路, 2024, 69(1): 270-275.
	LEI S D, LIN J, HUANG S L, et al. Research on improved YOLOv3 network for concrete surface crack detection[J]. Highway, 2024, 69(1): 270-275 (in Chinese).
[11]	王坤, 冯康威. 基于改进YOLOv5的交通场景小目标检测算法[J/OL]. 北京航空航天大学学报. (2024-03-12) [2024-04-05]. https://doi.org/10.13700/j.bh.1001-5965.2024.0003.
	WANG K, FENG K W. Small target detection algorithm for traffic scene based on improved YOLOv5[J/OL]. Journal of Beijing University of Aeronautics and Astronautics. (2024-03-12) [2024-04-05]. https://doi.org/10.13700/j.bh. 1001-5965.2024.0003 (in Chinese).
[12]	李利霞, 王鑫, 王军, 等. 基于特征融合与注意力机制的无人机图像小目标检测算法[J]. 图学学报, 2023, 44(4): 658-666. DOI
	LI L X, WANG X, WANG J, et al. Small object detection algorithm in UAV image based on feature fusion and attention mechanism[J]. Journal of Graphics, 2023, 44(4): 658-666 (in Chinese).
[13]	贾玉进, 张振程, 李浠铭, 等. 基于轻量化YOLOv5网络的输电线路绝缘子缺陷检测[J]. 电力学报, 2024, 39(1): 36-44.
	JIA J Y, ZHANG Z C, LI X M, et al. Defect detection of insulator on transmission line based on lightweight YOLOv5 network[J]. Journal of Electric Power, 2024, 39(1): 36-44 (in Chinese).
[14]	PENG D X, DING W, ZHEN T. A novel low light object detection method based on the YOLOv5 fusion feature enhancement[J]. Scientific Reports, 2024, 14(1): 4486. DOI PMID
[15]	PENG C, LI X Y, WANG Y L. TD-YOLOA: an efficient YOLO network with attention mechanism for tire defect detection[J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: 3529111.
[16]	朱强军, 胡斌, 汪慧兰, 等. 基于轻量化YOLOv8s交通标志的检测[J]. 图学学报, 2024, 45(3): 422-432. DOI
	ZHU Q J, HU B, WANG H L, et al. Detection of traffic signs based on lightweight YOLOv8s[J]. Journal of Graphics, 2024, 45(3): 422-432 (in Chinese). DOI
[17]	ZHANG Y L, WU Z J, WANG X, et al. Improved YOLOv8 insulator fault detection algorithm based on BiFormer[C]// 2023 IEEE 5th International Conference on Power, Intelligent Computing and Systems. New York: IEEE Press, 2023: 962-965.
[18]	DING X H, ZHANG X Y, HAN J G, et al. Diverse branch block: building a convolution as an inception-like unit[C]// 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2021: 10881-10890.
[19]	WANG C C, HE W, NIE Y, et al. Gold-YOLO: efficient object detector via gather-and-distribute mechanism[C]// The 37th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2023: 51094-51112.
[20]	YANG L X, ZHANG R Y, LI L D, et al. SimAM:a simple, parameter-free attention module for convolutional neural networks[C]// The 38th International Conference on Machine Learning. New York: PMLR, 2021: 11863-11874.
[21]	吴磊, 储钰昆, 杨洪刚, 等. 面向铝合金焊缝DR图像缺陷的Sim-YOLOv8目标检测模型[J]. 中国激光, 2024, 51(16): 1602103.1-1602103.10.
	WU L, CHU Y K, YANG H G, et al. Sim-YOLOv8 object detection model for DR image defects in aluminum alloy welds[J]. Chinese Journal of Lasers, 2024, 51(16): 1602103.1-1602103.10 (in Chinese).
[22]	魏陈浩, 杨睿, 刘振丙, 等. 具有双层路由注意力的YOLOv8道路场景目标检测方法[J]. 图学学报, 2023, 44(6): 1104-1111. DOI
	WEI C H, YANG R, LIU Z B, et al. YOLOv8 with bi-level routing attention for road scene object detection[J]. Journal of Graphics, 2023, 44(6): 1104-1111 (in Chinese).

基于DBBR-YOLO的光伏电池表面缺陷检测

Photovoltaic cell surface defect detection based on DBBR-YOLO

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 22

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李琼 , 考月英 , 张莹 , 徐沛 . 面向无人机航拍图像的目标检测研究综述[J]. 图学学报, 2024, 45(6): 1145-1164.
[2]	刘灿锋, 孙浩, 东辉. 结合 Transformer 与 Kolmogorov Arnold 网络的分子扩增时序预测研究[J]. 图学学报, 2024, 45(6): 1256-1265.
[3]	宋思程, 陈辰, 李晨辉, 王长波. 基于密度图多目标追踪的时空数据可视化[J]. 图学学报, 2024, 45(6): 1289-1300.
[4]	李珍峰, 符世琛, 徐乐, 孟博, 张昕, 秦建军. 基于MBI-YOLOv8的煤矸石目标检测算法研究[J]. 图学学报, 2024, 45(6): 1301-1312.
[5]	闫建红, 冉同霄. 基于YOLOv8的轻量化无人机图像目标检测算法[J]. 图学学报, 2024, 45(6): 1328-1337.
[6]	王宗继, 刘云飞, 陆峰. Cloud Sphere: 一种基于渐进式变形自编码的三维模型表征方法[J]. 图学学报, 2024, 45(6): 1375-1388.
[7]	许丹丹, 崔勇, 张世倩, 刘雨聪, 林予松. 优化医学影像三维渲染可视化效果：技术综述[J]. 图学学报, 2024, 45(5): 879-891.
[8]	胡凤阔, 叶兰, 谭显峰, 张钦展, 胡志新, 方清, 王磊, 满孝锋. 一种基于改进YOLOv8的轻量化路面病害检测算法[J]. 图学学报, 2024, 45(5): 892-900.
[9]	王亚茹, 冯利龙, 宋晓轲, 屈卓, 杨珂, 王乾铭, 翟永杰. TFD-YOLOv8：一种用于输电线路的异物检测方法[J]. 图学学报, 2024, 45(5): 901-912.
[10]	吴沛宸, 袁立宁, 胡皓, 刘钊, 郭放. 基于注意力特征融合的视频异常行为检测[J]. 图学学报, 2024, 45(5): 922-929.
[11]	翟永杰, 李佳蔚, 陈年昊, 王乾铭, 王新颖. 融合改进Transformer的车辆部件检测方法[J]. 图学学报, 2024, 45(5): 930-940.
[12]	刘丽, 张起凡, 白宇昂, 黄凯烨. 结合Swin Transformer的多尺度遥感图像变化检测研究[J]. 图学学报, 2024, 45(5): 941-956.
[13]	姜晓恒, 段金忠, 卢洋, 崔丽莎, 徐明亮. 融合先验知识推理的表面缺陷检测[J]. 图学学报, 2024, 45(5): 957-967.
[14]	章东平, 魏杨悦, 何数技, 徐云超, 胡海苗, 黄文君. 特征融合与层间传递：一种基于Anchor DETR改进的目标检测方法[J]. 图学学报, 2024, 45(5): 968-978.
[15]	李刚, 蔡泽浩, 孙华勋, 赵振兵. 基于改进YOLOv8与语义知识融合的金具缺陷检测方法研究[J]. 图学学报, 2024, 45(5): 979-986.