基于半监督学习的光伏板图像分割算法研究

doi:10.3778/j.issn.1002-8331.2404-0005

摘要/Abstract

摘要： 对光伏板的精确识别与分割是光伏产业近年来研究的一个热点话题。现有光伏板分割技术大多是基于语义分割来进行的，然而语义分割模型的精度往往依赖于数据集的规模，但在实际任务中获取大量精准标注的数据并非容易；同时光伏板图像数据存在着对比度不强、边界模糊、背景复杂等影响分割的问题，不能简单地将主流语义分割模型直接迁移到该任务上。因此，提出了一种基于半监督学习的光伏板分割算法。提出了一种基于FixMatch改进的半监督学习框架，使模型充分探索利用扰动空间的特征信息；设计了基于卷积神经网络（convolutional neural network，CNN）和Transformer组成的双分支特征聚合网络，在双分支主干网络的设计中充分利用了CNN和Transformer各自在局部特征和全局特征上提取的优势，并通过设计的特征聚合网络来将两条支路的特征信息进行充分聚合，从而使模型最大程度对原始图像的多级别特征信息进行学习；最后通过实验表明在0.1 m，0.3 m，0.8 m空间分辨率的屋顶分布式光伏数据集上，所提出的方法在仅使用1/32标注数据的情况下平均交并比（mean intersection over union，MIoU）指标分别可以达到83.74%、82.77%、80.73%。

关键词: 光伏产业, 语义分割, 半监督学习, 卷积神经网络, Transformer

Abstract: The precise identification and segmentation of photovoltaic panels have been a hot topic in the photovoltaic industry in recent years. Most existing photovoltaic panel segmentation technologies are based on semantic segmentation, but the accuracy of semantic segmentation models often depends on the size of the dataset. However, obtaining a large amount of accurately annotated data in practical tasks is not easy； at the same time, photovoltaic panel image data suffers from issues such as weak contrast, blurred boundaries, and complex backgrounds that affect segmentation, making it difficult to simply transfer mainstream semantic segmentation models directly to this task. Therefore, this article proposes a photovoltaic panel segmentation algorithm based on semi-supervised learning. Firstly, an improved semi-supervised learning framework based on FixMatch is proposed, which enables the model to fully explore and utilize the feature information of the perturbation space. Subsequently, a dual branch feature aggregation network based on convolutional neural network(CNN) and Transformer is designed. In the design of the dual branch backbone network, the advantages of CNN and Transformer in extracting local and global features are fully utilized, and the feature aggregation network is designed to fully aggregate the feature information of the two branches, thereby enabling the model to learn multi-level feature information of the original image to the greatest extent possible. Finally, experiments have shown that on the rooftop distributed photovoltaic dataset with spatial resolutions of 0.1 m, 0.3 m, and 0.8 m, the proposed method can achieve an average intersection over union (MIoU) index of 83.74%, 82.77%, and 80.73%, respectively, using only 1/32 annotated data.

Key words: photovoltaic industry, semantic segmentation, semi-supervised learning, convolutional neural networks, Transformer

苏俊, 刘智权, 唐潮龙. 基于半监督学习的光伏板图像分割算法研究[J]. 计算机工程与应用, 2025, 61(11): 325-334.

SU Jun, LIU Zhiquan, TANG Chaolong. Research on Photovoltaic Plate Image Segmentation Algorithm Based on Semi-Supervised Learning[J]. Computer Engineering and Applications, 2025, 61(11): 325-334.

参考文献

[1] LIU Z, DENG Z, HE G, et al. Challenges and opportunities for carbon neutrality in China[J]. Nature Reviews Earth and Environment, 2022, 3(2): 141-155.
[2] SACHS J D, SCHMIDT-TRAUB G, MAZZUCATO M, et al. Six transformations to achieve the sustainable development goals[J]. Nature Sustainability, 2019, 2(9): 805-814.
[3] SLAMERSAK A, KALLIS G, O’NEILL D W. Energy requirements and carbon emissions for a low-carbon energy transition[J]. Nature Communications, 2022, 13(1): 6932.
[4] SHUKLA P R, SKEA J, SLADE R, et al. Climate change 2022: mitigation of climate change[M]. Cambridge: Cambridge University Press, 2023.
[5] REN S M, HU W, BRADBURY K, et al. Automated extraction of energy systems information from remotely sensed data: a review and analysis[J]. Applied Energy, 2022, 326: 119876.
[6] JIE Y S, JI X H, YUE A Z, et al. Combined multi-layer feature fusion and edge detection method for distributed photovoltaic power station identification[J]. Energies, 2020, 13(24): 6742.
[7] 吴永静, 吴锦超, 林超, 等. 基于深度学习的高分辨率遥感影像光伏用地提取[J]. 测绘通报, 2021(5): 96-101.
WU Y J, WU J C, LIN C, et al. Photovoltaic land extraction from high-resolution remote sensing images based on deep learning method[J]. Bulletin of Surveying and Mapping, 2021(5): 96-101.
[8] 李青, 李海涛, 李辉, 等. 注意力机制和全局卷积在光伏板分割中的应用[J]. 计算机工程与应用, 2024, 60(4): 237-248.
LI Q, LI H T, LI H, et al. Photovoltaic panel segmentation using attention mechanism and global convolution[J]. Computer Engineering and Applications, 2024, 60(4): 237-248.
[9] CARION N, MASSA F, SYNNAEVE G, et al. End-to-end object detection with transformers[C]//Proceedings of the European Conference on Computer Vision. Cham: Springer International Publishing, 2020: 213-229.
[10] CHEN H T, WANG Y H, GUO T Y, et al. Pre-trained image processing transformer[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 12294-12305.
[11] DEVRIES T, TAYLOR G W. Improved regularization of convolutional neural networks with cutout[J]. arXiv:1708. 04552, 2017.
[12] ZHONG Z, ZHENG L, KANG G L, et al. Random erasing data augmentation[C]//Proceedings of the AAAI Conference on Artificial Intelligence, 2020: 13001-13008.
[13] SINGH K K, YU H, SARMASI A, et al. Hide-and-seek: a data augmentation technique for weakly-supervised locali- zation and beyond[J]. arXiv:1811.02545, 2018.
[14] CHEN P, LIU S, ZHAO H, et al. Gridmask data augmentation[J]. arXiv:2001.04086, 2020.
[15] CRESWELL A, WHITE T, DUMOULIN V, et al. Generative adversarial networks: an overview[J]. IEEE Signal Processing Magazine, 2018, 35(1): 53-65.
[16] MITTAL S, TATARCHENKO M, BROX T. Semi-supervised semantic segmentation with high-and low-level consistency[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 43(4): 1369-1379.
[17] CHEN X K, YUAN Y H, ZENG G, et al. Semi-supervised semantic segmentation with cross pseudo supervision[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 2613-2622.
[18] SOHN K, BERTHELOT D, LI C L, et al. FixMatch: simplifying semi-supervised learning with consistency and confidence[J]. arXiv:2001.07685, 2020.
[19] CARON M, MISRA I, MAIRAL J, et al. Unsupervised learning of visual features by contrasting cluster assignments[J]. arXiv:2006.09882, 2020.
[20] BERTHELOT D, CARLINI N, CUBUK E D, et al. Remixmatch: semi-supervised learning with distribution alignment and augmentation anchoring[J]. arXiv:1911.09785, 2019.
[21] ISLAM M A, JIA S, BRUCH N D B. How much position information do convolutional neural networks encode?[J]. arXiv:2001.08248, 2020.
[22] WANG Q L, WU B G, ZHU P F, et al. ECA-Net: efficient channel attention for deep convolutional neural networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 11531-11539.
[23] WU H P, XIAO B, CODELLA N, et al. CvT: introducing convolutions to vision transformers[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 22-31.
[24] JIANG H, YAO L, LU N, et al. Multi-resolution dataset for photovoltaic panel segmentation from satellite and aerial imagery[J]. Earth System Science Data, 2021, 13(11): 5389-5401.
[25] RONNEBERGER O, FISCHER P, BROX T. U-Net: convolutional networks for biomedical image segmentation[C]//Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham: Springer International Publishing, 2015: 234-241.
[26] ZHOU Z W, RAHMAN SIDDIQUEE M M, TAJBAKHSH N, et al. UNet++: a nested U-Net architecture for medical image segmentation[C]//Proceedings of the Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. Cham: Springer International Publishing, 2018: 3-11.
[27] CHEN L C, PAPANDREOU G, SCHROFF F, et al. Rethinking atrous convolution for semantic image segmentation[J]. arXiv:1706.05587, 2017.
[28] YU C, WANG J, PENG C, et al. BiSeNet: bilateral segmentation network for real-time semantic segmentation[C]//Proceedings of the European Conference on Computer Vision, 2018: 334-349.
[29] XIE E, WANG W, YU Z, et al. SegFormer: simple and efficient design for semantic segmentation with transformers[C]//Advances in Neural Information Processing Systems, 2021: 12077-12090.
[30] GU J Q, KWON H, WANG D L, et al. Multi-scale high-resolution vision transformer for semantic segmentation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 12084-12093.
[31] TARVAINEN A, VALPOLA H. Mean teachers are better role models: weight-averaged consistency targets improve semi-supervised deep learning results[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems, 2017: 1195-1204.
[32] YANG L H, ZHUO W, QI L, et al. ST++: make self-trainingwork better for semi-supervised semantic segmentation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2022: 4258-4267.