MGDE-UNet:轻量化光伏电池缺陷分割模型

doi:10.16088/j.issn.1001-6600.2024112202

摘要/Abstract

摘要： 针对光伏电池缺陷分割模型存在计算复杂度高、参数量大、分割速度慢和分割精度低的问题,本文提出一种基于轻量化改进U-Net的光伏电池缺陷分割模型。首先,使用MobitNetV3_Large网络替换原U-Net的主干网络,在减少模型计算量和参数量的同时,保留原网络的特征提取能力。其次,将DynamicConv模块融入GhostConv模块,设计出G-DConv模块,替换原U-Net上采样部分使用的普通卷积模块,在最大化减少网络参数和计算量的同时,提升模型的推理速度。最后,通过在网络上采样后引入ECA注意力机制,减少复杂背景对检测效果的干扰。实验结果表明,该模型的参数量仅为2.43×10⁶,计算量仅为3.03×10⁹,推理速度达到61 frame/s。相比基准模型,改进后的模型MIoU和MPA分别提升0.12个百分点和2.17个百分点,满足工业设备部署要求。

关键词: 光伏电池, U-Net, 轻量化, 语义分割, ECA

Abstract: Aiming at the problems of high computational complexity, large number of parameters, slow segmentation speed and low segmentation accuracy existing in the photovoltaic cell defect segmentation model, a photovoltaic cell defect segmentation model based on lightweight improved U-Net is proposed. First of all, the MobitNetv3_Large network is used to replace the backbone network of the original U-Net, which reduces the computational amount and the number of parameters of the model while retaining the feature extraction ability of the original network. Secondly, the G-DConv module is designed by integrating the DynamicConv module into the GhostConv module, replacing the ordinary convolutional module used in the upsampling part of the original U-Net, which maximally reduces the network parameters and computational amount while improving the inference speed of the model. Finally, by introducing the ECA attention mechanism after network upsampling, the interference of complex background on the detection effect is reduced. The experimental results show that the number of parameters of this model is only 2.43×10⁶, the computational amount is only 3.03×10⁹, and the inference speed reaches 61 frame/s. Compared with the baseline model, the improved model increases MIoU and MPA by 0.12 and 2.17 percentage points respectively, meeting the requirements for industrial equipment deployment.

Key words: photovoltaic cell, U-Net, light weight, semantic segmentation, ECA

中图分类号: TM914.4

王涛, 黎远松, 石睿, 陈慧宁, 侯宪庆. MGDE-UNet:轻量化光伏电池缺陷分割模型[J]. 广西师范大学学报（自然科学版）, 2026, 44(1): 45-55.

WANG Tao, LI Yuansong, SHI Rui, CHEN Huining, HOU Xianqing. MGDE-UNet: Defect Segmentation Model for Lightweight Photovoltaic Cells[J]. Journal of Guangxi Normal University(Natural Science Edition), 2026, 44(1): 45-55.

参考文献

[1] 刘亦丰. 基于机器视觉的太阳能电池缺陷检测方法及系统研究[D]. 合肥: 合肥工业大学, 2023. DOI: 10.27101/d.cnki.ghfgu.2023.002668.
[2] 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): 64-74. DOI: 10.12086/oee.2021.200112.
[3] 唐政, 张会林, 马立新. 密集主干网络算法对太阳能电池的缺陷检测[J]. 激光与光电子学进展, 2023, 60(14): 246-252. DOI: 10.3788/LOP222422.
[4] LU S, WU K X, CHEN J X. Solar cell surface defect detection based on optimized YOLOv5[J]. IEEE Access, 2023, 11: 71026-71036.
[5] CAO Y K, PANG D D, ZHAO Q C, et al. Improved YOLOv8-GD deep learning model for defect detection in electroluminescence images of solar photovoltaic modules[J]. Engineering Applications of Artificial Intelligence, 2024, 131: 107866. DOI: 10.1016/j.engappai.2024.107866.
[6] 郭建, 黄颖驹. 基于改进YOLOv5的光伏太阳能电池片缺陷检测[J]. 组合机床与自动化加工技术, 2024(11): 104-109. DOI: 10.13462/j.cnki.mmtamt.2024.11.020.
[7] 周颖, 颜毓泽, 陈海永, 等. 基于改进YOLOv8的光伏电池缺陷检测[J]. 激光与光电子学进展, 2024, 61(8): 245-255. DOI: 10.3788/LOP231622.
[8] 杨长春, 贺轩轩, 王睿, 等. 基于改进YOLOv8光伏板缺陷检测算法[J]. 电子测量技术, 2024, 47(23): 181-192. DOI: 10.19651/j.cnki.emt.2416770.
[9] PRATT L, GOVENDER D, KLEIN R. Defect detection and quantification in electroluminescence images of solar PV modules using U-Net semantic segmentation[J]. Renewable Energy, 2021, 178: 1211-1222. DOI: 10.1016/j.renene.2021.06.086.
[10] 王延年, 刘宏涛, 刘航宇, 等. 基于改进U-Net的太阳能电池图像缺陷检测方法[J]. 电子测量技术, 2021, 44(14): 117-121. DOI: 10.19651/j.cnki.emt.2106837.
[11] WANG C H, CHEN H Y, ZHAO S S, et al. Efficient and refined deep convolutional features network for the crack segmentation of solar cell electroluminescence images[J]. IEEE Transactions on Semiconductor Manufacturing, 2022, 35(4): 610-619. DOI: 10.1109/TSM.2022.3197933.
[12] 汪方斌, 李文豪. 基于改进轻量化U-Net模型的光伏电池EL图像缺陷检测[J]. 电子测量技术, 2024, 47(5): 102-111. DOI: 10.19651/j.cnki.emt.2415314.
[13] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention-MICCAI 2015. Cham: Springer, 2015: 234-241. DOI: 10.1007/978-3-319-24574-4_28.
[14] HOWARD A, SANDLER M, CHEN B, et al. Searching for MobileNetV3[C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV).Piscataway,NJ:IEEE, 2019: 1314-1324.. DOI: 10.1109/iccv.2019.00140.
[15] 王为, 赵涛, 钟羽中, 等. 基于SE-RetinaNet的面向玻璃面板的小尺寸低显著性缺陷检测[J]. 组合机床与自动化加工技术, 2024(7): 123-127, 131. DOI: 10.13462/j.cnki.mmtamt.2024.07.025.
[16] 宋益, 赵宁雨, 颜畅, 等. 隧道衬砌裂缝实时分割的Mobile-PSPNet方法[J]. 铁道科学与工程学报, 2022, 19(12): 3746-3757. DOI: 10.19713/j.cnki.43-1423/u.t20220024.
[17] 侯维岩, 靳东安, 王高杰, 等. 基于嵌入式系统的智能售货柜目标检测算法[J]. 电子测量与仪器学报, 2021, 35(10): 217-224. DOI: 10.13382/j.jemi.B2104002.
[18] 王维, 王显鹏, 宋相满. 基于自适应多目标进化CNN的图像分割方法[J]. 控制与决策, 2024, 39(4): 1185-1193. DOI: 10.13195/j.kzyjc.2022.1061.
[19] 陈法法, 董海飞, 何向阳, 等. 改进U_Net网络的钢结构表面锈蚀图像分割方法[J]. 电子测量与仪器学报, 2024, 38(2): 49-57. DOI: 10.13382/j.jemi.B2306742.
[20] 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).Piscataway,NJ: IEEE, 2020: 1577-1586. DOI: 10.1109/CVPR42600.2020.00165.
[21] 李忠科, 刘小芳. 基于轻量级YOLOv8n网络的PCB缺陷检测算法[J]. 电子测量技术, 2024, 47(4): 120-126. DOI: 10.19651/j.cnki.emt.2315076.
[22] HAN K, WANG Y H, GUO J Y, et al. ParameterNet: parameters are all you need for large-scale visual pretraining of mobile networks[C]//2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Piscataway,NJ: IEEE, 2024: 15751-15761. DOI: 10.1109/CVPR52733.2024.01491.
[23] OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]//ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).Piscataway,NJ:IEEE, 2023: 1-5. DOI: 10.1109/ICASSP49357.2023.10096516.
[24] 张萍, 徐巧枝. 基于多感受野与分组混合注意力机制的肺结节分割研究[J]. 广西师范大学学报(自然科学版), 2022, 40(3): 76-87. DOI: 10.16088/j.issn.1001-6600.2021070402.
[25] 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. DOI: 10.1109/TII.2022.3162846.
[26] 段小勇, 何超, 刘学渊. 基于改进DeepLabV3+的非结构化道路可行驶区域检测[J]. 中国农机化学报, 2025, 46(2): 271-278. DOI: 10.13733/j.jcam.issn.2095-5553.2025.02.040.
[27] 王福奇, 王志峰, 金建成, 等. 基于GSL-YOLO模型的综放工作面混矸率检测方法[J]. 工矿自动化, 2024, 50(9): 59-65, 137. DOI: 10.13272/j.issn.1671-251x.2024080011.
[28] 王海舰, 刘丽丽, 赵雪梅, 等. 基于改进PSPnet-MobileNetV2的煤岩界面快速精准识别[J]. 振动测试与诊断, 2024, 44(4): 793-800, 832-833. DOI: 10.16450/j.cnki.issn.1004-6801.2024.04.023.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed