Defect Detection of Photovoltaic Modules Based on Multi-Scale Feature Fusion

doi:10.3778/j.issn.1002-8331.2304-0390

Abstract

Abstract: A photovoltaic modules defect detection algorithm based on multi-scale feature fusion is proposed to address the challenges of complex defect backgrounds, large differences in defect scales, and a high number of small target defects that traditional object detection algorithms cannot solve. The algorithm is based on the YOLOv5s framework. Firstly, a coordinate attention mechanism is embedded in the backbone network to extract important defect shapes and enhance the network’s feature extraction ability. Secondly, a bidirectional feature pyramid network is used in the Neck network to adaptively fuse image features of different scales using adaptive weights. Finally, a tiny target detection layer is added to the prediction layer, and the ASFF detection head is used to adaptively fuse different output layers to reduce the loss of target feature information. The improved algorithm is validated on a photovoltaic component dataset, and the experimental results show that it can quickly and accurately identify defects, with an mAP of 91.9% and a recall rate of 90.8%, which represents a 3.2 and 4.5 percentage points improvement in mAP and recall rate, respectively, compared to the YOLOv5s network.

Key words: photovoltaic modules, YOLOv5, defect detection, feature fusion

摘要： 针对光伏组件缺陷背景复杂，缺陷尺度差异较大，小目标缺陷较多等传统目标检测算法无法解决的问题，提出了一种基于多尺度特征融合的光伏组件缺陷检测算法。算法以YOLOv5s为框架，在主干网络中嵌入坐标注意力机制，用于提取重要的缺陷形态，增强网络特征提取能力；在颈部网络中使用双向特征金字塔，以自适应权重的方式融合不同尺度的图像特征；在预测层添加微小目标检测层，并结合ASFF检测头自适应融合不同输出层，减少目标特征信息丢失。在光伏组件数据集上进行验证，实验表明改进后的算法可以快速精准识别缺陷，其中mAP达到了91.9%，召回率达到了90.8%，相比于YOLOv5s网络，mAP和召回率分别提升了3.2、4.5个百分点。

关键词: 光伏组件, YOLOv5, 缺陷检测, 特征融合

TIAN Hao, ZHOU Qiang, HE Chenlong. Defect Detection of Photovoltaic Modules Based on Multi-Scale Feature Fusion[J]. Computer Engineering and Applications, 2024, 60(3): 340-347.

田浩, 周强, 贺晨龙. 基于多尺度特征融合的光伏组件缺陷检测[J]. 计算机工程与应用, 2024, 60(3): 340-347.

References

[1] 苏健, 梁英波, 丁麟, 等. 碳中和目标下我国能源发展战略探讨[J]. 中国科学院院刊, 2021, 36(9): 1001-1009.
SU J, LIANG Y B, DING L, et al. Research on China’s energy development strategy under carbon neutrality[J]. Bulletin of Chinese Academy of Sciences, 2021, 36(9): 1001-1009.
[2] PARIDA B, INIYAN S, GOIC R. A review of solar photovoltaic technologies[J]. Renewable and Sustainable Energy Reviews, 2011, 15(3): 1625-1636.
[3] DHIMISH M, HOLMES V, MEHRDADI B, et al. The impact of cracks on photovoltaic power performance[J]. Journal of Science: Advanced Materials and Devices, 2017, 2(2): 199-209.
[4] HENISCH H K. Electroluminescence[J]. Reports on Progress in Physics, 1964, 27(1): 369.
[5] MANSOURI A, ZETTL M, MAYER O, et al. Defect detection in photovoltaic modules using electroluminescence imaging[C]//27th European Photovoltaic Solar Energy Conference and Exhibition, 2012: 3374-3378.
[6] FUYUKI T, KITIYANAN A. Photographic diagnosis of crystalline silicon solar cells utilizing electroluminescence[J]. Applied Physics A, 2009, 96: 189-196.
[7] SU B, CHEN H, CHEN P, et al. Deep learning-based solar-cell manufacturing defect detection with complementary attention network[J]. IEEE Transactions on Industrial Informatics, 2020, 17(6): 4084-4095.
[8] CHEN Y, DING Y, ZHAO F, et al. Surface defect detection methods for industrial products: a review[J]. Applied Sciences, 2021, 11(16): 7657.
[9] TSAI D M, WU S C, LI W C. Defect detection of solar cells in electroluminescence images using Fourier image reconstruction[J]. Solar Energy Materials and Solar Cells, 2012, 99: 250-262.
[10] ANWAR S A, ABDULLAH M Z. Micro-crack detection of multicrystalline solar cells featuring an improved anisotropic diffusion filter and image segmentation technique[J]. EURASIP Journal on Image and Video Processing, 2014, 2014: 1-17.
[11] TSAI D M, WU S C, CHIU W Y. Defect detection in solar modules using ICA basis images[J]. IEEE Transactions on Industrial Informatics, 2012, 9(1): 122-131.
[12] 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.
[13] TANG W, YANG Q, XIONG K, et al. Deep learning based automatic defect identification of photovoltaic module using electroluminescence images[J]. Solar Energy, 2020, 201: 453-460.
[14] HAN H, GAO C, ZHAO Y, et al. Polycrystalline silicon wafer defect segmentation based on deep convolutional neural networks[J]. Pattern Recognition Letters, 2020, 130: 234-241.
[15] LI X, YANG Q, LOU Z, et al. Deep learning based module defect analysis for large-scale photovoltaic farms[J]. IEEE Transactions on Energy Conversion, 2018, 34(1): 520-529.
[16] HOU Q, ZHOU D, FENG J. Coordinate attention for efficient mobile network design[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13713-13722.
[17] TAN M, PANG R, LE Q V. Efficientdet: scalable and efficient object detection[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 10781-10790.
[18] LIU S, HUANG D, WANG Y. Learning spatial fusion for single-shot object detection[J]. arXiv:1911.09516, 2019.
[19] LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 8759-8768.
[20] SU B, ZHOU Z, CHEN H. PVEL-AD: a large-scale open-world dataset for photovoltaic cell anomaly detection[J]. IEEE Transactions on Industrial Informatics, 2022, 19(1): 404-413.
[21] BUERHOP-LUTZ C, DEITSCH S, MAIER A, et al. A benchmark for visual identification of defective solar cells in electroluminescence imagery[C]//35th European PV Solar Energy Conference and Exhibition, 2018: 1287-1289.
[22] DEITSCH S, BUERHOP-LUTZ C, SOVETKIN E, et al. Segmentation of photovoltaic module cells in uncalibrated electroluminescence images[J]. Machine Vision and Applications, 2021, 32(4): 84.
[23] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
[24] WOO S, PARK J, LEE J Y, et al. Cbam: convolutional block attention module[C]//Proceedings of the European Conference on Computer Vision (ECCV), 2018: 3-19.
[25] LIU Y, SHAO Z, HOFFMANN N. Global attention mechanism: retain information to enhance channel-spatial interactions[J]. arXiv:2112.05561, 2021.