通过无人机拍摄光伏组件红外图像进行检测时,图像背景较为复杂,且其中含有的小目标热斑故障在检测过程中容易丢失信息,出现误检或漏检等状况。针对上述问题,将HCF-Net与YOLOv8n网络结合,提出一种融合网络（HCF-YOLO）用于光伏组件红外图像热斑故障检测。加入并行化注意力机制（PPA）,通过分层特征融合和注意力机制来增强小目标的表达,确保在多个降采样步骤后保留热斑关键信息。采用维度感知选择性集成模块（DASI）,注重对高维和低维特征的自适应选择和精细融合,增强小目标的显著性。使用PIoU作为HCF-YOLO的损失函数,在回归的早期阶段,引导预测框沿有效路径回归,提升检测速度。HCF-YOLO算法相较于原有的YOLOv8n算法,检测精度（AP50）从89.27%提升至97.28%,并且检测速度达到217.33帧/s,实验结果可证明模型的有效性。
Abstract
When using drones to capture infrared images of photovoltaic modules for detection, the images often exhibit complex backgrounds, leading to challenges in detecting small target hot spots, which may result in information loss, false positives, or false negatives. In response to these issues, a novel fusion network（HCF-YOLO） is proposed by integrating the HCF-Net with the YOLOv8n network for detecting hot spot faults in infrared images of photovoltaic modules. This fusion network incorporates a Parallelized Patch-Aware Attention Module（PPA） to enhance the representation of small targets through hierarchical feature fusion and attention mechanisms, ensuring the retention of crucial hot spot information after multiple downsampling steps. Additionally, a Dimension-Aware Selective Integration Module（DASI） is employed to adaptively select and finely integrate high-dimensional and low-dimensional features, thereby enhancing the saliency of small targets. The Powerful IoU（PIOU） is adopted as the loss function for HCF-YOLO, guiding the early-stage regression of predicted boxes along effective paths to improve detection speed. The HCF-YOLO algorithm demonstrates a significant enhancement in detection accuracy, with the average precision at an intersection over union of 50%（AP50） increasing from 89.27% to 97.28% compared to the original YOLOv8n algorithm, achieving a detection speed of 217.33 frame/s. According to the experimental results, the effectiveness of the model is validated.
关键词
光伏组件 /
红外图像 /
目标检测 /
YOLOv8 /
深度学习 /
注意力机制
Key words
PV modules /
infrared imaging /
target detection /
YOLOv8 /
deep learning /
attention mechanism
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 王道累, 姚从荣, 李超, 等. 光伏组件热斑效应智能化检测的综述及展望[J]. 太阳能学报, 2024, 45(5): 527-536.
WANG D L, YAO C R, LI C, et al.Overview and prospect of intelligent detection of hot spot effect of photovoltaic modules[J]. Acta energiae solaris sinica, 2024, 45(5): 527-536.
[2] ZUNIGA-REYES M A, ROBLES-OCAMPO J B, SEVILLA-CAMACHO P Y, et al. Photovoltaic failure detection based on string-inverter voltage and current signals[J]. IEEE access, 2021, 9: 39939-39954.
[3] 耿洪彬, 张英杰, 魏燕飞, 等. 基于多源数据的光伏电站在线热斑检测性能评价系统研究[J]. 山东电力技术, 2023, 50(1): 14-19.
GENG H B, ZHANG Y J, WEI Y F, et al.Online hot spot detection performance evaluation system for photovoltaic power station based on multi-source data[J]. Shandong electric power, 2023, 50(1): 14-19.
[4] 刘卫亮, 姜锴越, 许之胜, 等. 基于数字孪生与融合神经网络的光伏阵列故障诊断[J].太阳能学报,2024, 45(11): 303-312.
LIU W L, JIANG K Y, XU Z S, et al.Fault diagnosis of photovoltaic array based on digital twin and fusion neural network[J]. Acta energiae solaris sinica, 2024, 45(11): 303-312.
[5] 贾帅康, 白英君, 孙海蓉, 等. 基于DenseNet的红外图像热斑状态分类研究[J]. 山东电力技术, 2021, 48(3): 60-64.
JIA S K, BAI Y J, SUN H R, et al.Research on classification of hot spot state in infrared image based on dense net[J]. Shandong electric power, 2021, 48(3): 60-64.
[6] 陈辉, 张傲, 孙帅, 等. 基于改进DeepLabv3+的含噪声热红外图像光伏热斑检测方法[J]. 太阳能学报, 2023, 44(11): 23-30.
CHEN H, ZHANG A, SUN S, et al.Photovoltaic thermal spot detection method with noisy thermal infrared image based on improved DeepLabv3+[J]. Acta energiae solaris sinica, 2023, 44(11): 23-30.
[7] 王银, 沈灵鑫, 李茂环, 等. 融合视觉特征的光伏组件语义分割模型研究[J]. 太阳能学报, 2024, 45(4): 500-511.
WANG Y, SHEN L X, LI M H, et al.Semantic segmentation model of photovoltaic modules based on visual features[J]. Acta energiae solaris sinica, 2024, 45(4): 500-511.
[8] 李琼, 吴文宝, 刘斌, 等. 基于迁移学习的光伏组件鸟粪覆盖检测[J]. 太阳能学报, 2022, 43(2): 233-237.
LI Q, WU W B, LIU B, et al.Bird droppings coverage detection of photovoltaic module based on transfer learning[J]. Acta energiae solaris sinica, 2022, 43(2): 233-237.
[9] GIRSHICK R.Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 2015: 1440-1448.
[10] WANG F B, WANG Z N, CHEN Z, et al.An edge-guided deep learning solar panel hotspot thermal image segmentation algorithm[J]. Applied sciences, 2023, 13(19): 11031.
[11] GAO X F, WANG T, LIU M M, et al.A framework to identify guano on photovoltaic modules in offshore floating photovoltaic power plants[J]. Solar energy, 2024, 274: 112598.
[12] REN Y F, YU Y J, LI J, et al.Design of photovoltaic hot spot detection system based on deep learning[J]. Journal of physics: conference series, 2020, 1693(1): 012075.
[13] LIU H, DONG Z Y.ESA-SSD: single-stage object detection network using deep hierarchical feature learning[J]. Multimedia tools and applications, 2024, 83(18): 56207-56228.
[14] 艾上美, 周剑峰, 张必朝, 等. 基于改进SSD算法的光伏组件缺陷检测研究[J]. 智慧电力, 2023, 51(12): 53-58.
AI S M, ZHOU J F, ZHANG B C, et al.Defect detection of photovoltaic modules based on improved SSD algorithm[J]. Smart power, 2023, 51(12): 53-58.
[15] YU J, GUAN R Q, ZHANG C G, et al.A novel object recognition method for photovoltaic (PV) panel occlusion based on deep learning[J]. Journal of computational methods in sciences and engineering, 2023, 23(6): 3391-3405.
[16] WANG Y, ZHAO J Y, XIE G, et al.PA-YOLO-based multifault defect detection algorithm for PV panels[J]. International journal of photoenergy, 2024, 2024: 6113260.
[17] XU S B, ZHENG S C, XU W H, et al.HCF-Net: Hierarchical context fusion network for infrared small object detection[C]//2024 IEEE International Conference on Multimedia and Expo(ICME). Niagara Falls, ON, Canada, 2024.
[18] ZHENG Z H, WANG P, LIU W, et al.Distance-IoU loss: faster and better learning for bounding box regression[J]. Proceedings of the AAAI conference on artificial intelligence, 2020, 34(7): 12993-13000.
[19] LIU C, WANG K G, LI Q, et al.Powerful-IoU: More straightforward and faster bounding box regression loss with a nonmonotonic focusing mechanism[J]. Neural networks, 2024, 170: 276-284.
[20] WENG K H, CHU X X, XU X M, et al. Efficientrep: an efficient Repvgg-style ConvNets with hardware-aware neural network design[J]. arXiv preprint arXiv:2302.00386, 2023.
[21] DING X H, ZHANG X Y, MA N N, et al.RepVGG: making VGG-style ConvNets great again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA, 2021: 13733-13742.
[22] GEVORGYAN Z. SIoU loss: More powerful learning for bounding box regression[J]. arXiv preprint arXiv:2205.12740, 2022.
[23] REZATOFIGHI H, TSOI N, GWAK J, et al.Generalized intersection over union: a metric and a loss for bounding box regression[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Long Beach, CA, USA, 2019: 658-666.
[24] TONG Z J, CHEN Y H, XU Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[EB/OL].2023: 2301.10051. https://arxiv.org/abs/2301.10051v3.
PDF(1598 KB)
