浏览全部资源
扫码关注微信
福建省计量科学研究院 国家光伏产业计量测试中心,福建 福州 350003福建省计量科学研究院 福建省能源计量重点实验室,福建 福州 350003
[ "何翔(1990-),男,福建福州人,硕士,工程师,2016年于中南大学获得硕士学位,主要从事从光伏计量方面研究工作。E-mail:315107763@qq.com" ]
网络出版日期:2023-08-23,
收稿日期:2023-07-04,
修回日期:2023-07-31,
扫 描 看 全 文
何翔, 杨爱军, 黎健生, 等. 基于cycleGAN的太阳电池电致发光图像数据增强方法[J/OL]. 液晶与显示, 2023,1-14.
HE xiang, YANG aijun, LI jiansheng, et al. Electroluminescence defect image augmentation method of solar cell based on cycleGAN[J/OL]. Chinese Journal of Liquid Crystals and Displays, 2023,1-14.
何翔, 杨爱军, 黎健生, 等. 基于cycleGAN的太阳电池电致发光图像数据增强方法[J/OL]. 液晶与显示, 2023,1-14. DOI: 10.37188/CJLCD.2023-0234.
HE xiang, YANG aijun, LI jiansheng, et al. Electroluminescence defect image augmentation method of solar cell based on cycleGAN[J/OL]. Chinese Journal of Liquid Crystals and Displays, 2023,1-14. DOI: 10.37188/CJLCD.2023-0234.
为了解决光伏组件电致发光缺陷自动识别研究中训练用图像不足以及生成图像质量不佳的问题,采用cycleGAN生成了太阳电池EL缺陷图像,并且将生成的图像与具有代表性的DCGAN所生成的图像进行了对比。将拍摄到的EL图像进行分类、进行数据增强以形成训练集。接着采用训练集训练两个模型。最后,从生成图像的有效性,相似性,多样性三个角度对比了两个模型所生成图像。cycleGAN所生成的图像中,有效图像的占比显著高于DCGAN;与真实拍摄的图像相比,cycleGAN所生成的图像感官相似性极高,不存在 “脏污感”和“粗糙感”,难以通过人眼分辨。在FID指标上,cycleGAN所生成的图像显著低于DCGAN;采用cycleGAN生成图像训练的分类模型中,对真实拍摄的图像组成的测试集准确率达到93.45%,当训练集中混入少量真实拍摄的图像时,该准确率提升至到98.26%,显著高于DCGAN;cycleGAN生成图像的平均MS-SSIM指标显著低于DCGAN。采用cycleGAN进行太阳电池EL图像的数据增强是一种行之有效的方法。在有效性,相似性,多样性三个面显著优于DCGAN。
In order to solve the problem of insufficient training images and poor quality of generated images in the automatic recognition research of electroluminescence defects (EL) in photovoltaic modules, The solar cell EL defect images are generated by using the cycleGAN, and the generated images is compared with the images generated by the representative DCGAN. Classify the EL images and perform data augmentation to form a training set. Next, cycleGAN and DCGAN were trained using training set. Finally, a detailed comparison was made between the generated images of the two models from three perspectives: effectiveness, similarity, and diversity. The experimental results show that the proportion of effective images generated by cycleGAN is significantly higher than images generated by DCGAN. Compared with captured EL images, the images generated by cycleGAN have extremely high sensory similarity, making it difficult to distinguish them through the human eye. And there is no “dirty” or “rough” phenomenon. For FID indicators, the images generated by cycleGAN are significantly lower than those generated by DCGAN. The classification model trained with images generated by cycleGAN achieved a 93.45% accuracy rate on the test set composed of captured EL images. When a small number of captured EL images were included in the training dataset, the accuracy is improved to 98.26%, significantly higher than that of DCGAN. Finally, the average MS-SSIM indicators of images generated by cycleGAN is significantly lower than that of DCGAN. Using cycleGAN for data augmentation of solar cell EL images is an effective method. And it performs significantly better than DCGAN in terms of effectiveness, similarity, and diversity.
光伏组件太阳电池电致发光cycleGANDCGAN
Photovoltaics moduleSolar cellsElectroluminescencecycleGANDCGAN
林剑春,杨爱军,沈熠辉.电致发光缺陷检测仪的成像性能评估[J].光学精密工程,2017,25(6):1418-1424. doi: 10.3788/ope.20172506.1418http://dx.doi.org/10.3788/ope.20172506.1418
LIN J C, YANG A J, SHEN Y H. Evaluation of imaging performance for electroluminescence defect detector[J]. Optics and Precision Engineering, 2017,25(6):1418-1424. (in Chinese). doi: 10.3788/ope.20172506.1418http://dx.doi.org/10.3788/ope.20172506.1418
刘斌辉,黄璇,杨强荣,等. 基于GoogLeNet与EL的光伏电池缺陷检测技术研究[J]. 电子产品可靠性与环境试验,2019(0):150-155.
LIU B H, HUANG X, YANG Q R, et al.. Research on defect detection technology of photovoltaic cells based on GoogLeNet and EL[J]. Electronic Product Reliability and Environmental Testing, 2019(0):150-155. (in Chinese)
那峙雄,樊涛,孙涛,等.多损失融合的小样本光伏组件隐裂检测算法[J].计算机科学与探索,2022,16(2):458-467. doi: 10.3778/j.issn.1673-9418.2111036http://dx.doi.org/10.3778/j.issn.1673-9418.2111036
NA Z X, FAN T, SU T, et al.. Micro-cracks detection of solar cells based on few shot samples with multi-loss[J]. Journal of Frontiers of Computer Science and Technology,2022,16(2):458-467. (in Chinese). doi: 10.3778/j.issn.1673-9418.2111036http://dx.doi.org/10.3778/j.issn.1673-9418.2111036
赵晓雨,高林,杨校李,等. 融合注意力机制的YOLOv5光伏板电致发光图像缺陷检测算法[J]. 湖北民族大学学报(自然科学版),2023,41(1):65-70.
ZHAO X Y, GAO L, YANNG X L, et al.. Algorithm of YOLOv5 fusing attention mechanism for defect detection of photovoltaic panel electroluminescent image[J]. Journal of Hubei Minzu University(Natural Science Edition), 2023,41(1):65-70. (in Chinese)
沈凌云,张洁,孙甲云,等.基于改进Alexnet模型的光伏电池片瑕疵分类方法[J].自动化与信息工程,2022,43(4):42-47. doi: 10.3969/j.issn.1674-2605.2022.04.009http://dx.doi.org/10.3969/j.issn.1674-2605.2022.04.009
SHEN L Y, ZHANG J, SUN J Y, et al.. Defect classification method of photovoltaic cells based on Improved alexnet model[J]. Automation & Information Engineering, 2022,43(4):42-47. (in Chinese). doi: 10.3969/j.issn.1674-2605.2022.04.009http://dx.doi.org/10.3969/j.issn.1674-2605.2022.04.009
周建凯,许盛之,赵二刚,等. 基于深度学习的电池片缺陷识别研究[J].电子技术应用,2019,45(5):66-69,77.
ZHOU J K, XU S Z, ZHAO E G, et al.. Cell defect recognition based on deep learning[J]. Application of Electronic Technique, 2019,45(5):66-69,77. (in Chinese)
吴涛,赖菲.基于LeNet-5模型的太阳能电池板缺陷识别分类[J].热力发电,2019,48(3):120-125. doi: 10.19666/j.rlfd.201807147http://dx.doi.org/10.19666/j.rlfd.201807147
WU T, LAI F. Identification and classification of defects in solar cells based on LeNet-5 model[J]. Thermal Power Generation, 2019,45(5):66-69,77. (in Chinese). doi: 10.19666/j.rlfd.201807147http://dx.doi.org/10.19666/j.rlfd.201807147
沈凌云,朱明,陈小云.基于径向基神经网络的太阳能电池缺陷检测[J].发光学报,2015,36(1):99-105. doi: 10.3788/fgxb20153601.0099http://dx.doi.org/10.3788/fgxb20153601.0099
SHEN L Y, CHEN X Y. Solar panels defect detection based on radial basis function neural network[J]. Chinese Journal of Luminescence, 2015,36(1):99-105. (in Chinese). doi: 10.3788/fgxb20153601.0099http://dx.doi.org/10.3788/fgxb20153601.0099
刘怀广,丁晚成,黄千稳.基于轻量化卷积神经网络的光伏电池片缺陷检测方法研究[J].应用光学,2022,43(1):87-94.
LIU H G, DING W C, HUANG. Defects detection method of photovoltaic cells based on lightweight convolutional neural network[J]. Journal of Applied Optics, 2022,43(1):87-94. (in Chinese)
墨恺,徐林.基于阈值均匀局部二值模式和BP神经网络的太阳电池缺陷检测算法[J].太阳能学报,2014,35(12):2448-2454. doi: 10.3969/j.issn.0254-0096.2014.12.018http://dx.doi.org/10.3969/j.issn.0254-0096.2014.12.018
MO K, XU L. Defect detection of solar cell based on threshold uniform bp and bp neural network[J]. Acta Energiae Solaris Sinica, 2014,35(12):2448-2454. (in Chinese). doi: 10.3969/j.issn.0254-0096.2014.12.018http://dx.doi.org/10.3969/j.issn.0254-0096.2014.12.018
陈海永,赵鹏,闫皓炜.融合注意力的多尺度Faster RCNN的裂纹检测[J].光电工程,2021,48(1):61-71.
CHEN H Y, ZHAO P, YAN H W. Crack detection based on multi-scale Faster RCNN with attention[J]. Opto-Electronic Engineering,2014,35(12):2448-2454. (in Chinese)
文熙. 基于生成对抗网络的太阳能电池缺陷数据增强与缺陷检测[D]. 河北:河北工业大学,2020.
WEN X. Data augmentation and defect detection of solar cells based on generative adversarial network[D]. Hebei University of Technology,2020. (in Chinese)
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. (in Chinese)
张一平,许盛之,孟子尧,等. 基于FC-ACGAN网络的太阳电池EL图像的数据增强方法[J]. 太阳能学报,2021,42(10):35-41.
ZHANG Y P, XU S Z, MENG Z Y, et al.. Data enhancement method of solar cell el image based on FC-ACGAN network[J]. Acta Energiae Solaris Sinica,2021,42(10):35-41. (in Chinese)
周颖,叶红,王彤,等. 基于多尺度CNN的光伏组件缺陷识别[J]. 太阳能学报,2022,43(2):211-216.
ZHOUY, YE H, WANG T, et al.. Photovoltaic module defect identification based on mulit-scale convolution neural network[J]. Acta Energiae Solaris Sinica, 2022,43(2):211-216. (in Chinese)
刘坤,文熙,黄闽茗,等. 基于生成对抗网络的太阳能电池缺陷增强方法[J]. 浙江大学学报(工学版),2020,54(4):684-693.
LIU K, WEN X, HUANG M M, et al.. Solar cell defect enhancement method based on generative adversarial network[J]. Journal of Zhejiang University (Engineering Science) ,2020,54(4):684-693. (in Chinese)
GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C]. Proceedings of the 27th International Conference on Neural Information Processing Systems, Montreal, Cambridge: MIT Press, 2014: 2672-2680.
何翔. 基于DCGANs的半片光伏组件电致发光图像增强技术[J]. 应用光学,2023,44(2):314-322. doi: 10.5768/jao202344.0202003http://dx.doi.org/10.5768/jao202344.0202003
HE X. Electroluminescence image enhancement technology of half-cut photovoltaic module based on DCGANsP[J]. Journal of Applied Optics,2023,44(2):314-322. (in Chinese). doi: 10.5768/jao202344.0202003http://dx.doi.org/10.5768/jao202344.0202003
梁俊杰, 韦舰晶, 蒋正锋. 生成对抗网络GAN综述[J]. 计算机科学与探索, 2020, 14(1): 1-17.
LIANG J J, WEI J J, JIANG Z F. Generative adversarial networks GAN overview[J], Journal of Frontiers of Computer Science & Technology, 2020, 14(1): 1-17. (in Chinese)
RADFORD A, METZ L, CHINTALA S, et al.. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv:1511.06434, 2015. doi: 10.1109/aiar.2018.8769811http://dx.doi.org/10.1109/aiar.2018.8769811
ZHU J Y, PARK T, ISOLA PET AL. Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks[C]. 2017 IEEE International Conference on Computer Vision 2017, Venice, Italy, 2017:2242-2251. doi: 10.1109/iccv.2017.244http://dx.doi.org/10.1109/iccv.2017.244
ULYANOV D, VEDALDI A, VLEMPITSKYet al.. Instance Normalization: The Missing Ingredient for Fast Stylization[J]. ArXiv,2016:1607.08022. doi: 10.1109/cvpr.2017.437http://dx.doi.org/10.1109/cvpr.2017.437
HEUSEL M, RAMSAUER H, TUNTERTHINERe alt.. GANs trained by a two time-scale update rule converge to a local NASH equilibrium[C]. Advances in Neural Information Processing Systems 2017. Long Beach,California USA, 2017:6626-6637. doi: 10.48550/arXiv.1706.08500http://dx.doi.org/10.48550/arXiv.1706.08500
WANG Z, SIMONCELLI E P, BOVIK A C. Multi-scale structural similarity for image quality assessment[C]. Proceedings of the 37th IEEE Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, 2003: 1398-1402.
0
浏览量
7
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构