针对太阳电池EL图像,提出一种高识别率的轻量化M-CNN网络模型。首先,该网络模型将拼接的多特征图融合通道注意力机制;然后,引入Ghost卷积层降低模型参数;最后,利用普通卷积层取代最大池化层进行特征空间降维。实验结果表明：在自建裂痕、阴影、微小瑕疵、无缺陷图像数据库共15767片上,M-CNN对粗糙分类检测准确率和瑕疵分类检测准确率分别是99.83%和93.38%,模型参数量是1.29 MB。相较先进的MobileNetV3、DeepVit和MobileVit等网络,M-CNN有缺陷识别率高和模型参数量低的优势。
Abstract
A lightweight micro convolutional neural network （M-CNN） model with high recognition rate is proposed for EL images of solar cells. The network model incorporates a fusion channel attention mechanism to merge multiple feature maps. Introducing Ghost convolutional layers to reduce the model parameters, and using ordinary convolutional layers to replace maximum pooling layers for feature space dimensionality reduction. Experimental results show that on a self-built database of 15767 EL images of cracks, shadows, minor defects, and no defects, M-CNN achieves an accuracy of 99.83% and 93.38% for rough classification and flaw classification detection, respectively, with a model parameter count of 1.29 MB. Notably, compared to advanced networks such as MobileNetV3, DeepVit, and MobileVit, M-CNN has the advantages of superior defect recognition rate and lower model parameter count.
关键词
太阳电池 /
深度学习 /
卷积神经网络 /
图像处理 /
缺陷识别
Key words
solar cells /
deep learning /
convolutional neural networks /
image processing /
defect identification
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] LI G Q, AKRAM M W, JIN Y, et al.Thermo-mechanical behavior assessment of smart wire connected and busbarPV modules during production, transportation, and subsequent field loading stages[J]. Energy, 2019, 168: 931-945.
[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] TSAI D M, LI G N, LI W C, et al.Defect detection in multi-crystal solar cells using clustering with uniformity measures[J]. Advanced engineering informatics, 2015, 29(3): 419-430.
[4] 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 algorithm of solar cell based on threshold uniform local binary model and BP neural network[J]. Acta energiae solaris sinica, 2014, 35(12): 2448-2454.
[5] GE C P, LIU Z, FANG L M, et al.A hybrid fuzzy convolutional neural network based mechanism for photovoltaic cell defect detection with electroluminescence images[J]. IEEE transactions on parallel and distributed systems, 2021, 32(7): 1653-1664.
[6] XU C, FAMOURI M, BATHLA G, et al.CellDefectNet: a machine-designed attention condenser network for electroluminescence-based photovoltaic cell defect inspection[C]//2022 19th Conference on Robots and Vision (CRV). Toronto, ON, Canada, 2022: 219-223.
[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] 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.
[9] WOO S, PARK J, LEE J Y, et al.CBAM: convolutional block attention module[C]//European Conference on Computer Vision. Cham: Springer, 2018: 3-19.
[10] HAN K, WANG Y H, TIAN Q, et al.GhostNet: more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA, 2020: 1577-1586.
[11] HOWARD A, SANDLER M, CHEN B, et al.Searching for MobileNetV3[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV). Seoul, Korea (South), 2019: 1314-1324.
[12] HU J, SHEN L, SUN G.Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA, 2018: 7132-7141.
[13] HE K M, ZHANG X Y, REN S Q, et al.Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas, NV, USA, 2016: 770-778.
[14] SZEGEDY C, VANHOUCKE V, IOFFE S, et al.Rethinking the inception architecture for computer vision[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Las Vegas, NV, USA, 2016: 2818-2826.
[15] YUAN L, CHEN Y P, WANG T, et al.Tokens-to-token ViT: training vision transformers from scratch on ImageNet[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV). Montreal, QC, Canada, 2021: 538-547.
[16] HAN K, XIAO A, WU E, et al.Transformer in transformer[J]. Advances in neural information processing systems, 2021, 34: 15908-15919.
[17] ZHOU D Q, KANG B Y, JIN X J, et al. DeepViT: towards deeper vision transformer[EB/OL].2021: arXiv: 2103.11886. http://arxiv.org/abs/2103.11886.
[18] MEHTA S, RASTEGARI M. MobileViT: light-weight, general-purpose,mobile-friendly vision transformer[EB/OL].2021: arXiv: 2110.02178. http://arxiv.org/abs/2110.02178.
[19] HUANG G, LIU Z, VAN DER MAATEN L, et al. Densely connected convolutional networks[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA, 2017: 2261-2269.
[20] TAN M X, LE Q V. EfficientNet: rethinking model scaling for convolutional neural networks[EB/OL].2019: arXiv: 1905.11946. http://arxiv.org/abs/1905.11946.
基金
辽宁省教育厅研究生教改项目六位一体、产教融合的专业学位硕士研究生实践创新能力培养模式探索（LNYJG2023117）
PDF(2871 KB)
