提出并设计名为OD-YOLO缺陷检测模型来改善多晶硅太阳电池电致发光成像中复杂背景干扰缺陷检测效果的问题。使用二次卷积模块（TwiceConv-OD）过滤掉复杂晶粒背景干扰,增强模型对缺陷本身的关注力;提出anchor-plus1分配策略来增加模型在面对复杂背景时获取更多的缺陷正样本数量,提升模型的准确率与召回率,减少漏检误检;使用K-means++算法初始化锚框尺寸,聚类后的锚框更能代表检测样本中所有缺陷的几何形貌,能更好适应多晶硅太阳电池的缺陷尺度差异。经公开缺陷检测数据集PVEL-AD-2021的实验验证：OD-YOLO模型的均值平均精度（mAP）达到89.4%,相比YOLOV5s缺陷检测模型提升3%,准确率提升4.8%,召回率提升1.9%,参数减少4.5%,检测速度为104帧/s。
Abstract
A defect detection model named OD-YOLO is proposed and designed to improve the problem of complex background interfering with the defect detection effect in electroluminescence imaging of polycrystalline silicon solar cells. We use the secondary convolution module （TwiceConv-OD） to filter out the complex grain background interference and enhance the model's focus on the defects themselves; we propose the anchor-plus1 allocation strategy to increase the number of defect positive samples obtained by the model in the face of the complex background, which improves the model's accuracy and recall and reduces the omission of misdetections; and we use the K-means++ algorithm to initialize the anchor frames and cluster the anchor frames to be more representative of all the detected samples. The size of the anchor frame is initialized using the K-means++ algorithm, and the clustered anchor frame is more representative of the geometrical shape of all defects in the detected samples, which is better adapted to the differences in defect scales of polysilicon solar cells. Experimentally verified by the publicly available defect detection dataset PVEL-AD-2021： the mean average precision （mAP） of the OD-YOLO model reaches 89.4%, which is improved by 3% compared to the YOLOV5s defect detection model, the accuracy is improved by 4.8%, the recall rate is improved by 1.9%, the parameters are reduced by 4.5%, and the speed is 104 frames per second.
关键词
太阳电池 /
计算机视觉 /
表面缺陷 /
卷积神经网络
Key words
solar cells /
computer vision /
surface defects /
convolutional neural network
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] OSTAPENKO S, DALLAS W, HESS D, et al.Crack detection and analyses using resonance ultrasonic vibrations in crystalline silicon wafers[C]//2006 IEEE 4th World Conference on Photovoltaic Energy Conference. Waikoloa, HI, USA, 2006: 920-923.
[2] 范程华, 王群京, 曹欣远, 等. 基于欠定方程的太阳电池片表面缺陷检测算法[J]. 太阳能学报, 2020, 41(6): 288-292.
FAN C H, WANG Q J, CAO X Y, et al.Defect detection algorithm for solar cells surface based on underdetermined equation[J]. Acta energiae solaris sinica, 2020, 41(6): 288-292.
[3] YANG W R.Short-time discrete wavelet transform for wafer microcrack detection[C]//2009 IEEE International Symposium on Industrial Electronics. Seoul, Korea (South), 2009: 2069-2074.
[4] KÖNTGES M, MORLIER A, EDER G, et al. Review: ultraviolet fluorescence as assessment tool for photovoltaic modules[J]. IEEE journal of photovoltaics, 2020, 10(2): 616-633.
[5] SCHLOTHAUER J, JUNGWIRTH S, RODER B, et al.Fluorescence imaging: a powerful tool for the investigation of polymer degradation in PV modules[J]. Journal of econometrics, 2010, 10: 149-154.
[6] SCHLOTHAUER J, JUNGWIRTH S, KÖHL M, et al. Degradation of the encapsulant polymer in outdoor weathered photovoltaic modules: spatially resolved inspection of EVA ageing by fluorescence and correlation to electroluminescence[J]. Solar energy materials and solar cells, 2012, 102: 75-85.
[7] LYU Y, FAIRBROTHER A, KIM J, et al.Fluorescence imaging analysis of depth-dependent degradation in photovoltaic laminates: insights to the failure[J]. Progress in photovoltaics: research and applications, 2020, 28(2): 122-134.
[8] HE Y Z, DU B L, HUANG S D.Noncontact electromagnetic induction excited infrared thermography for photovoltaic cells and modules inspection[J]. IEEE transactions on industrial informatics, 2018, 14(12): 5585-5593.
[9] MENG L, MA F J, WONG J, et al.Extraction of surface recombination velocity at highly doped silicon surfaces using electron-beam-induced current[J]. IEEE journal of photovoltaics, 2015, 5(1): 263-268.
[10] GEISTHARDT R M, SITES J R.Nonuniformity characterization of CdTe solar cells using LBIC[J]. IEEE journal of photovoltaics, 2014, 4(4): 1114-1118.
[11] DI TOMMASO A, BETTI A, FONTANELLI G, et al.A multi-stage model based on YOLOv3 for defect detection in PV panels based on IR and visible imaging by unmanned aerial vehicle[J]. Renewable energy, 2022, 193: 941-962.
[12] HAN H, GAO C Q, ZHAO Y, et al.Polycrystalline silicon wafer defect segmentation based on deep convolutional neural networks[J]. Pattern recognition letters, 2020, 130: 234-241.
[13] DEITSCH S, CHRISTLEIN V, BERGER S, et al.Automatic classification of defective photovoltaic module cells in electroluminescence images[J]. Solar energy, 2019, 185: 455-468.
[14] RAHMAN M R U, CHEN H Y. Defects inspection in polycrystalline solar cells electroluminescence images using deep learning[J]. IEEE access, 2020, 8: 40547-40558.
[15] SU B Y, CHEN H Y, CHEN P, et al.Deep learning-based solar-cell manufacturing defect detection with complementary attention network[J]. IEEE transactions on industrial informatics, 2021, 17(6): 4084-4095.
[16] 陶志勇, 杜福廷, 任晓奎, 等. 基于T-VGG的太阳电池片缺陷检测[J]. 太阳能学报, 2022, 43(7): 145-151.
TAO Z Y, DU F T, REN X K, et al.Defect detection of solar cells based on T-VGG[J]. Acta energiae solaris sinica, 2022, 43(7): 145-151.
[17] 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.
MO K, XU L.Defect detection of solar cell based on threshold uniform LBP and BP neural network[J]. Acta energiae solaris sinica, 2014, 35(12): 2448-2454.
[18] LI C, ZHOU A, YAO A. Omni-dimensional dynamic convolution[J]. arXiv preprint arXiv:2209. 07947, 2022.
[19] ZHU X Z, HU H, LIN S, et al.Deformable convnets v2: More deformable, better results[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 9308-9316.
[20] 鲁伟明, 李省, 张付特, 等. 基于不同电压下的电致发光和红外成像的太阳电池缺陷检测[J]. 发光学报, 2014, 35(12): 1511-1519.
LU W M, LI X, ZHANG F T, et al.Defect detection of solar cell based on electroluminescence and thermography imaging with different bias voltage[J]. Chinese journal of luminescence, 2014, 35(12): 1511-1519.
[21] TSAI D M, WU S C, CHIU W Y.Defect detection in solar modules using ICA basis images[J]. IEEE transactions on industrial informatics, 2013, 9(1): 122-131.
[22] SU B Y, ZHOU Z, CHEN H Y.PVEL-AD: a large-scale open-world dataset for photovoltaic cell anomaly detection[J]. IEEE transactions on industrial informatics, 2023, 19(1): 404-413.
[23] GLENN JOCHER.YOLOv5[EB/OL].https://github.com/ultralytics.
[24] GE Z, LIU S, WANG F, et al.Yolox: Exceeding yolo series in 2021[J]. arXiv preprint arXiv: 2107.08430, 2021.
[25] 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[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 7464-7475.
[26] JOCHER G, CHAURASIA A, QIU J. YOLOv8 by Ultralytics[CP/OL]. (2023-01)[2023-06-27]. https://github.com/ultralytics/ultralytics.
[27] LIU W, ANGUELOV D, ERHAN D, et al.Ssd: Single shot multibox detector[C]//Computer Vision-ECCV 2016: 14th European Conference. Amsterdam, The Netherlands, 2016: 21-37.
基金
湖南省自然科学基金（2021JJ30717）
PDF(4714 KB)
