由于航拍光伏红外图像中的热斑故障多为小目标且与干扰背景极为相似,导致热斑故障检测精度低,基于此提出基于深度学习的二阶段式热斑检测方法。第一阶段,进行干扰背景去除。针对光伏组件分割速度慢、边缘提取效果较差和正负样本不均衡的问题,通过替换主干网络和采用混合损失函数,提出一种改进的DeepLabv3+分割模型,实现光伏组件区域的快速、精准提取;第二阶段,进行热斑故障检测。针对小目标热斑漏检、误检问题,通过采用增强版SPP模块、引入浅层检测尺度和改变边框回归损失函数,提出一种改进的YOLOv5热斑检测模型,实现热斑的准确识别。利用自制数据集开展对比试验,结果表明相比于原DeepLabv3+分割模型,所提分割模型的平均像素准确率和平均交并比分别提高1.7和1.51个百分点;相比于原YOLOv5模型,所提热斑检测模型的平均精度均值mAP50与mAP50∶95分别提高2.6和10.7个百分点。
Abstract
Heat spot faults in aerial photovoltaic infrared images are mostly small targets, which are very similar to the interference background, resulting in low detection accuracy of heat spot faults. A second-order segment heat spot detection method based on deep learning is proposed. In the first stage, interference background removal is performed. Aiming at the problems of slow segmentation speed, poor edge extraction effect and unbalanced positive and negative samples of photovoltaic modules, an improved DeepLabv3+ segmentation model was proposed by replacing the backbone network and adopting hybrid loss function to achieve fast and accurate extraction of photovoltaic module regions. In the second stage, the hot spot fault detection is carried out. To solve the problem of missing and false detection of small target hot spots, an improved YOLOv5 hot spot detection model was proposed by adopting the enhanced SPP module, introducing shallow detection scale and changing the border regression loss function, so as to achieve accurate identification of hot spots. The results show that compared with the original DeepLabv3+ segmentation model, the average pixel accuracy and average crossover ratio of the proposed segmentation model are increased by 1.7 and 1.51 percentage points, respectively. Compared with the original YOLOv5 model, the average accuracy of mAP50 and mAP50∶95 of the proposed heat spot detection model is increased by 2.6 and 10.7 percentage paints, respectively.
关键词
光伏组件 /
红外图像 /
热斑 /
深度学习 /
DeepLabv3+ /
YOLOv5
Key words
PV modules /
infrared imagery /
hot spot /
deep learning /
DeepLabv3+ /
YOLOv5
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 水电水利规划设计总院. 中国可再生能源发展报告-2022[M]. 北京: 中国水利水电出版社, 2023.
Hydropower and Water Resources Planning and Design Institute. China renewable energy development report(2022)[M]. Beijing: China Water & Power Press, 2023.
[2] 杨亚楠. 太阳能光伏阵列识别及热斑检测技术的研究与实现[D]. 南京: 南京邮电大学, 2018.
YANG Y N.Design and implementation of solar photovoltaic array identification and hot spot detection technology[D]. Nanjing: Nanjing University of Posts and Telecommunications, 2018.
[3] 郭梦浩, 徐红伟. 基于Faster RCNN的红外热图像热斑缺陷检测研究[J]. 计算机系统应用, 2019, 28(11): 265-270.
GUO M H, XU H W.Hot spot defect detection based on infrared thermal image and Faster RCNN[J]. Computer systems & applications, 2019, 28(11): 265-270.
[4] 孙海蓉, 李号. 基于深度迁移学习的小样本光伏热斑识别方法[J]. 太阳能学报, 2022, 43(1): 406-411.
SUN H R, LI H.Photovoltaic hot spot identification method for small sample based on deep transfer learning[J]. Acta energiae solaris sinica, 2022, 43(1): 406-411.
[5] GERD IMENES A, SAAD NOORI N, ANDREAS NESVÅG UTHAUG O, et al. A deep learning approach for automated fault detection on solar modules using image composites[C]//2021 IEEE 48th Photovoltaic Specialists Conference (PVSC), Fort Lauderdale, FL, USA, 2021: 1925-1930.
[6] HONG F, SONG J, MENG H, et al.A novel framework on intelligent detection for module defects of PV plant combining the visible and infrared images[J]. Solar energy, 2022, 236: 406-416.
[7] 王道累, 李超, 李明山, 等. 基于深度卷积神经网络的光伏组件热斑检测[J]. 太阳能学报, 2022, 43(1): 412-417.
WANG D L, LI C, LI M S, et al.Solar photovoltaic modules hot spot detection based on deep convolutional neural networks[J]. Acta energiae solaris sinica, 2022, 43(1): 412-417.
[8] 王道累, 李明山, 姚勇, 等. 改进SSD的光伏组件热斑缺陷检测方法[J]. 太阳能学报, 2023, 44(4): 420-425.
WANG D L, LI M S, YAO Y, et al.Method of hotspot detection of photovoltaic panels modules on improved SSD[J]. Acta energiae solaris sinica, 2023, 44(4): 420-425.
[9] CHEN L C, ZHU Y K, PAPANDREOU G, et al.Encoder-decoder with atrous separable convolution for semantic image segmentation[M]. Cham: Springer International Publishing, 2018: 833-851.
[10] MA Y D, LIU Q, QIAN Z B.Automated image segmentation using improved PCNN model based on cross-entropy[C]//Proceedings of 2004 International Symposium on Intelligent Multimedia, Video and Speech Processing. Hong Kong, China, 2004: 743-746.
[11] SUDRE C H, LI W, VERCAUTEREN T, et al.Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations[C]// International Workshop on Deep Learning in Medical Image Analysis,Quebec, Canada,2017.
[12] LIN T Y, DOLLAR P, GIRSHICK R, et al.Feature Pyramid Networks for Object Detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI, USA, 21-26 July 2017.
[13] LIU S, QI L, QIN H, et al.Path aggregation network for instance segmentation[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake, USA, 2018.
[14] 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.
[15] HE K M, ZHANG X Y, REN S Q, et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.
[16] WANG C Y, MARK LIAO H Y, WU Y H, et al. CSPNet: a new backbone that can enhance learning capability of CNN[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops(CVPRW). Seattle, WA, USA, 2020: 1571-1580.
[17] 李小军, 邓月明, 陈正浩, 等. 改进YOLOv5的机场跑道异物目标检测算法[J]. 计算机工程与应用, 2023, 59(2): 202-211.
LI X J, DENG Y M, CHEN Z H, et al.Improved YOLOv5’s foreign object debris detection algorithm for airport runways[J]. Computer engineering and applications, 2023, 59(2): 202-211.
基金
国家能源集团科技项目（GJNY-21-98）;中央高校基本科研业务费（2023JG005;2023JC010）
PDF(2601 KB)
