作者：彭兴辉, 田建平, 吴相东, 黄浩平, 鞠杰
作者单位：四川轻化工大学机械工程学院，四川 宜宾 644005
关键词：电池片;图像处理;生成对抗网络;目标检测网络
摘要：
为准确快速检测和定位太阳能电池硅片的缺陷，判别缺陷种类。以太阳能隐裂缺陷片为研究对象，利用形态学方法进行图像预处理，通过改进的生成对抗网络（P256-BEGAN）生成图像数据，并利用FID评估其生成图像的质量，解决数据集不充足导致目标检测网络准确率低的问题；采用YOLOv5目标检测网络，对其训练过程及推理过程进行改进，实现太阳能电池片缺陷的检测及定位。试验结果表明，生成图像数据作为改进的目标检测网络训练集，准确率可达94.1%，单张电池片检测时间最短可达9 ms；混合真假数据之后，准确率可提高3.1%，满足工业电池片实时检测需求。
Research on defect detection method of photovoltaic cell based on data enhancement
PENG Xinghui, TIAN Jianping, WU Xiangdong, HUANG Haoping, JU Jie
School of Mechanical Engineering, Sichuan University of Science & Engneering, Yibin 644005, China
Abstract: In order to accurately and quickly detect and locate the defects of solar cell silicon wafer and distinguish the types of defects. Taking solar hidden crack defect slice as the research object, morphological method was used to preprocess the image. An improved generation countermeasure network (P256-began) was used to generate image data, and FID was used to evaluate the quality of the generated image. The problem of low accuracy of target detection network caused by insufficient data set was solved. The YOLOv5 target detection network is used to improve the training process and reasoning process, and the detection and location of solar cell defects are realized. The experimental results show that the generated image data can be used as the improved target detection network training set, and the accuracy can reach 94.1%, and the detection time of single battery can reach 9 ms. When real and fake data are mixed, the accuracy is improved by 3.1%. To achieve the industrial battery real-time detection requirements.
Keywords: cell;image processing;generating adversarial network;target detection network
2023, 49(7):29-34 收稿日期: 2021-10-16;收到修改稿日期: 2021-12-03
基金项目: 四川省科技厅重点研发项目（2022YFS0552）
作者简介: 彭兴辉(1995-),男,四川达州市人,硕士研究生,专业方向为机器视觉、深度学习
参考文献
[1] 程家早. 太阳能发电系统的组成[J]. 大众用电, 2015, 30(3): 23
[2] 赵思童. 光伏新能源技术在建筑设计中的应用[J]. 电池, 2023, 53(1): 125-126
[3] 殷鹰, 谢罗峰, 黄泰博. 基于深度学习的磁瓦内部缺陷声振检测方法[J]. 中国测试, 2020, 46(3): 32-38
[4] 刘桂雄, 何彬媛, 吴俊芳, 等. 面向视觉检测的深度学习图像分类网络及在零部件质量检测中应用[J]. 中国测试, 2019, 45(7): 1-10
[5] TAKAHASHI Y, KAJI Y, OGANE A, et al. "Luminoscopy": Novel tool for the diagnosis of crystalline silicon solar cells and modules utilizing electroluminescence[C]// 2006 IEEE 4th World Conference on Photovoltaic Energy Conference. IEEE, 2007.
[6] 邓堡元, 何赟泽, 王洪金, 等. 光伏电池图像序列的深度学习检测方法[J]. 机械工程学报, 2021, 57(8): 98-106
[7] 伊欣同, 单亚峰. 基于改进Faster R-CNN的光伏电池内部缺陷检测[J]. 电子测量与仪器学报, 2021, 35(1): 40-47
[8] SUN Q, MELNIKOV A, MANDELIS A. Camera-based high frequency heterodyne lock-in carrierographic (frequency-domain photoluminescence) imaging of crystalline silicon wafers[J]. Physica Status Solidi A, 2015, 213(2): 405-411
[9] 杨晓君, 任现坤, 韩玉杰, 等. 抗隐裂光伏组件的研究与应用[J]. 山东科学, 2020, 33(6): 126-130
[10] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]// International Conference on Neural Information Processing Systems. MIT Press, 2014: 2672-2680.
[11] RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[EB/OL]. (2015-11-15)[2020-10-13]. https://www.researchgate.net/publication/284476553_ Unsupervised_ Representation_ Learning_ with_ Deep_ Convolutional_ Generative_ Adversarial_ Networks.
[12] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein generative adversarial networks[C]//International Conference on Machine Learning. PMLR, 2017: 214-223.
[13] BERTHELOT D, SCHUMM T, METZ L. BEGAN: Boundary equilibrium generative adversarial networks[EB/OL]. [2017-5-31].https://arxiv.org/abs/1703.10717v4
[14] SALIMANS T, GOODFELLOW I, ZAREMBA W, et al. Improved techniques for training GANs[J]. Advances in Neural Information Processing Systems. Cambridge: MIT Press, 2016: 2234-2242
[15] XU Q, HUANG G, YUAN Y, et al. An empirical study on evaluation metrics of generative adversarial networks[DB/OL]. [2018-06-19]. https://arxiv.org/abs/1806.07755.
[16] REDMON J, FARHADI A. YOLO9000: Better, faster, stronger[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2017: 6517-6525.
