针对太阳电池缺陷种类复杂多样、人工检测效率低下的问题,提出一种基于迁移学习与ResNet的太阳电池缺陷检测方法。该方法首先利用残差模块搭建ResNet缺陷检测模型,接着将参数从ImageNet预训练模型迁移到ResNet缺陷检测模型中并使用电致发光（EL）图像微调模型,然后加载最优模型构建缺陷检测分类器实现太阳电池的缺陷检测。将该方法应用到常州某企业太阳电池的缺陷检测中。结果表明：在不同数据比例下,所提方法综合性能最好、泛化能力强、鲁棒性好,能实现模型快速训练并提高太阳电池缺陷检测的准确率,可满足太阳电池缺陷检测的需求。
Abstract
In view of the complex and diverse types of solar cell defects and low manual detection efficiency, this paper proposes a solar cell defect detection method based on transfer learning and ResNet. This method first builds the ResNet defect detection model with the residual module, then uses the ImageNet pre-training model to initialize the ResNet model, and uses the EL image to fine-tune the defect detection model. Finally, the optimal model is loaded to build a classifier to realize the defect detection of solar cells. The proposed method uses transfer learning to optimize the training process, which can realize rapid model training and improve the accuracy of defect detection in the condition of small samples. To verify the effectiveness of the proposed method, both the proposed approach and some conventional approaches were applied to the real dataset collected from a solar cell production line in Changzhou. The comparative experiments show that the proposed method has the best comprehensive performance, strong generalization ability, and robustness. The proposed method can effectively realize the defect detection of solar cells and improves the production efficiency and quality of solar cells.
关键词
迁移学习 /
深度学习 /
太阳电池 /
缺陷检测 /
EL图像
Key words
transfer learning /
deep learning /
solar cells /
defect detection /
EL images
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] SAMPAIO P G V, GONZÁLEZ M O A. Photovoltaic solar energy: conceptual framework[J]. Renewable and sustainable energy reviews, 2017, 74: 590-601.
[2] 杨博, 赵勇, 谢小军, 等. PERC电池EL黑斑对于光伏组件性能的影响[J]. 太阳能学报, 2020, 41(12): 97-102.
YANG B, ZHAO Y, XIE X J, et al.Effect of EL dark spot on PERC photovoltaic module performance[J]. Acta energiae solaris sinica, 2020, 41(12): 97-102.
[3] 王宪保, 李洁, 姚明海, 等. 基于深度学习的太阳能电池片表面缺陷检测方法[J]. 模式识别与人工智能, 2014, 27(6): 517-523.
WANG X B, L I J, YAN M H, et al. Solar cells surface defects detection based on deep learning[J]. Pattern recognition and artificial intelligence, 2014, 27(6): 517-523.
[4] SUN M, LYU S, ZHAO X, et al.Defect detection of photovoltaic modules based on convolutional neural network[C]//Machine Learning and Intelligent Communications: Second International Conference(MLICOM), Weihai, China, 2018.
[5] 周颖, 毛立, 张燕, 等. 改进CNN的太阳电池缺陷识别方法研究[J]. 太阳能学报, 2020, 41(12): 69-76.
ZHOU Y, MAO L, ZHANG Y, et al.Research on defect detection and classification for solar cells based on improved convolutional neural network[J]. Acta energiae solaris sinica, 2020, 41(12): 69-76.
[6] AKEAM M W, LI G, JIN Y, et al.CNN based automatic detection of photovoltaic cell defects in electroluminescence images[J]. Energy, 2019, 189: 116319.
[7] TANG W Q, YANG Q, XIONG K X, et al.Deep learning based automatic defect identification of photovoltaic module using electroluminescence images[J]. Solar energy, 2020, 201: 453-460.
[8] 龙明盛. 迁移学习问题与方法研究[D]. 北京: 清华大学, 2014.
LONG M S.Transfer learning problems and methods[D]. Beijing: Tsinghua University, 2014.
[9] WANG M, DENG W.Deep visual domain adaptation: a survey[J]. Neurocomputing, 2018, 312: 135-153.
[10] PAN S J, YANG Q.A survey on transfer learning[J]. IEEE transactions on knowledge and data engineering, 2010, 22(10): 1345-1359.
[11] MADADI Y, SEYDI V, NASROLLAHI K, et al.Deep visual unsupervised domain adaptation for classification tasks: a survey[J]. IET image processing, 2020, 14(14): 3283-3299.
[12] 罗维平, 徐洋, 陈永恒, 等. 基于迁移学习和改进ResNet50网络的织物疵点检测算法[J]. 毛纺科技, 2021, 49(2): 71-78.
LUO W P, XU Y, CHEN Y H, et al.Fabric defect detection algorithm based on migration learning and improved of ResNet50 network[J]. Wool textile journal, 2021, 49(2): 71-78.
[13] HE K M, ZHANG X Y, REN S Q, et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, NV, USA, 2016.
[14] GLOROT X, BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[C]//Proceedings of the thirteenth International Conference on Artificial Intelligence and Statistics(AISTATS), Sardinia, Italy, 2010.
基金
国家重点研发计划（2020YFB1711201）
PDF(2081 KB)
