基于Real-ESRGAN和改进YOLOv8n的输电线路绝缘子故障检测

doi:10.3969/j.issn.2097-0706.2024.07.004

摘要/Abstract

摘要：

为解决无人机在输电线路巡检时遇到的绝缘子故障难以检测的问题，提出一种绝缘子故障检测新方法。该方法结合了真实世界增强超分辨率生成对抗网络（Real-ESRGAN）和改进的YOLOv8n。首先，利用Real-ESRGAN对数据集进行超分辨率重构，优化数据集质量，有效减少复杂背景的干扰；然后利用高效视觉变压器框架替换YOLOv8的主干，加强模型的特征提取能力，同时使模型在推理阶段有更快的处理速度；再对YOLOv8的检测头进行轻量化处理，进一步加速模型推理。试验结果显示，该方法的均值平均精度达86.7%，证明了其在复杂背景下的卓越目标检测性能。通过分析热力图，展示了该算法与传统YOLOv8在关注区域上的差异，从而揭示了模型的内部工作机理。

关键词: 目标检测, 输电线路, 绝缘子, 无人机, YOLOv8, 超分辨重构, 生成对抗网络

Abstract:

To solve the difficulties in insulator fault detection during drone-based transmission line inspections， an innovative method for insulator fault detection which combines Real Enhanced Super Resolution Generative Adversarial Network （Real-ESRGAN） and improved YOLOv8n is proposed. Firstly， Real-ESRGAN is adopted to perform super-resolution reconstruction on the dataset， in order to optimize the quality of the dataset and effectively reduce the interference of complex backgrounds. Then， an efficient visual transformer framework is used to replace the backbone of YOLOv8， enhancing the model's feature extraction ability and accelerating the image process at inference stage. Lightweight processing is carried out on detection heads of YOLOv8 to further accelerate the inference of the model. Experimental results show that the mean average precision of this detection method reaches 86.7%， demonstrating its excellent object detection performance in complex backgrounds. Finally， by analyzing heat maps， the differences between the results made by the proposed method and the traditional YOLOv8 in the focus area are displayed， revealing the internal working mechanism of the mode.

Key words: object detection, transmission line, insulator, drone, YOLOv8, super-resolution reconstruction, generative adversarial network

中图分类号:

TM721

任一鸣, 杜董生, 邓祥帅, 连贺, 赵哲敏. 基于Real-ESRGAN和改进YOLOv8n的输电线路绝缘子故障检测[J]. 综合智慧能源, 2024, 46(7): 29-39.

REN Yiming, DU Dongsheng, DENG Xiangshuai, LIAN He, ZHAO Zhemin. Fault detection for transmission line insulators based on Real-ESRGAN and improved YOLOv8n[J]. Integrated Intelligent Energy, 2024, 46(7): 29-39.

图/表 18

图1

图2

图3

图4

图5

图6

图7

图8

表1

图9

图10

图11

图12

表2

表3

图13

图14

图15

参考文献 26

[1]	赵春宇, 吴红石, 熊志群, 等. 高压输电线路的防雷技术[J]. 模具制造, 2023, 23(11): 175-177.
	ZHAO Chunyu, WU Hongshi, XIONG Zhiqun, et al. Lightning protection technology for high-voltage transmission lines[J]. Mold Manufacturing, 2023, 23(11): 175-177.
[2]	赵鑫, 钱本华, 王睿, 等. 电化学储能参与电网安全稳定控制的研究综述[J]. 综合智慧能源, 2023, 45(1):58-66. doi: 10.3969/j.issn.2097-0706.2023.01.007
	ZHAO Xin, QIAN Benhua, WANG Rui, et al. A review of research on the participation of electrochemical energy storage in power grid safety and stability control[J]. Integrated Intelligent Energy, 2023, 45(1): 58-66. doi: 10.3969/j.issn.2097-0706.2023.01.007
[3]	王江, 薛铁柱, 魏龙. 无人机盘煤在火力发电企业燃料管理中的应用[J]. 华电技术, 2019, 41(6):78-80.
	WANG Jiang, XUE Tiezhu, WEI Long. Application of unmannedaerial vehicle coal mining infuel management of thermal power generation enterprises[J]. Huadian Technology, 2019, 41(6):78-80.
[4]	杨瑞鹏. 输电线路无人机巡检系统对绝缘子检测的前沿技术[J]. 电器工业, 2023(11):65-67.
	YANG Ruipeng. The cutting-edge technology of insulator detection in unmanned aerial vehicle inspection systems for transmission lines[J]. Electrical Industry, 2023(11):65-67.
[5]	肖新帅, 田秀霞, 徐曼. 基于端到端算法的绝缘子检测技术研究[J]. 华电技术, 2021, 43(2):28-33.
	XIAO Xinshuai, TIAN Xiuxia, XU Man. Research on insulator detection technology based on end-to-end algorithm[J]. Huadian Technology, 2021, 43(2):28-33.
[6]	NAWAZ S A, LI J B, BHATTI U A, et al. AI-based object detection latest trends in remote sensing, multimedia and agriculture applications[J]. Frontiers in Plant Science, 2022, 13: 1041514.
[7]	SHETTY A K, SAHA I, SANGHVI R M, et al. A review: Object detection models[C]// International Conference for Convergence in Technology (I2CT),IEEE, 2021: 1-8.
[8]	VASHISHT M, KUMAR B. A survey paper on object detection methods in image processing[C]// International Conference on Computer Science,Engineering and Applications (ICCSEA),IEEE, 2020: 1-4.
[9]	崔得东, 冯涛, 白昆, 等. 复杂背景目标探测识别技术综述[J]. 前瞻科技, 2022, 1(4): 69-80. doi: 10.3981/j.issn.2097-0781.2022.04.005
	CUI Dedong, FENG Tao, BAI Kun, et al. Overview of complex background target detection and recognition technology[J]. Forward Technology, 2022, 1(4): 69-80.
[10]	刘颖, 刘红燕, 范九伦, 等. 基于深度学习的小目标检测研究与应用综述[J]. 电子学报, 2020, 48(3): 590-601. doi: 10.3969/j.issn.0372-2112.2020.03.024
	LIU Ying, LIU Hongyan, FAN Jiulun, et al. A review of research and application on small object detection based on deep learning[J]. Journal of Electronics, 2020, 48 (3):590-601.
[11]	ZHOU X, JIANG L, HU C X, et al. YOLO-SASE: An improved YOLO algorithm for the small targets detection in complex backgrounds[J]. Sensors, 2022, 22(12):4600-4614.
[12]	SUN C Y, CHEN Y J, XIAO C, et al. YOLOv5s-DSD: An improved aerial image detection algorithm based on YOLOv5s[J]. Sensors(Basel), 2023, 23(15):6905-6920.
[13]	LIN Z W, HUANG M, ZHOU Q H. Infrared small target detection based on YOLOv4[C]// Journal of Physics: Conference Series, 2023, 2450(1): 012019.
[14]	ZHAO L, ZHI L Q, ZHAO C, et al. Fire-YOLO: A small target object detection method for fire inspection[J]. Sustainability, 2022, 14(9): 4930.
[15]	PARK S C, PARK M K, KANG M G. Super-resolution image reconstruction: a technical overview[J]. IEEE Signal Processing Magazine, 2003, 20(3):21-36.
[16]	ZHANG J, SHAO M H, YU L L, et al. Image super-resolution reconstruction based on sparse representation and deep learning[J]. Signal Processing: Image Communication, 2020, 87:115925.
[17]	GU J J, LU H N, ZUO W M, et al. Blind super-resolution with iterative kernel correction[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 1604-1613.
[18]	GRAVES D, PEDRYCZ W. Kernel-based fuzzy clustering and fuzzy clustering: A comparative experimental study[J]. Fuzzy Sets and Systems, 2010, 161(4): 522-543.
[19]	WANG X T, YU K, WU S X, et al. Esrgan: Enhanced super-resolution generative adversarial networks[C]// Proceedings of the European Conference on Computer Vision (ECCV) Workshops, 2018:1-16.
[20]	JANG D W, PARK R H. DenseNet with deep residual channel-attention blocks for single image super resolution[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019: 1-9.
[21]	SCHONFELD E, SCHIELE B, KHOREVA A. A U-Net based discriminator for generative adversarial networks[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020:8207-8216.
[22]	MUTHUKUMAR V, NARANG A, SUBRAMANIAN V, et al. Classification vs regression in overparameterized regimes: Does the loss function matter?[J]. Journal of Machine Learning Research, 2021, 22(222): 1-69.
[23]	LU X K, MA C, NI B B, et al. Deep regression tracking with shrinkage loss[C]// Proceedings of the European Conference on Computer vision (ECCV), 2018:353-369.
[24]	LIU X Y, PENG H W, ZHENG N X, et al. EfficientViT: Memory efficient vision transformer with cascaded group attention[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023:14420-14430.
[25]	WANG Y, LI Y, TONG H H, et al. HIT: Nested named entity recognition via head-tail pair and token interaction[C]// Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing(EMNLP), 2020: 6027-6036.
[26]	HORVÁTH A, HILLMER M, LOU Q, et al. Cellular neural network friendly convolutional neural networks——CNNs with CNNs[C]// Design,Automation & Test in Europe Conference & Exhibition (DATE),IEEE, 2017: 145-150.

配置名称	版本参数
显卡	RTX 4090(24 GB)
处理器	22vCPU AMD EPYC 7T83 64-Core Processor
操作系统	AutoDL AI
Touch	2.0.0+cu118
Opencv	4.8.1.78
Ultralytics	8.0.209

绝缘子故障	P/%	R/%	m_AP@0.5
Two Hight Glass	82.8	97.7	0.905
Glass Dirty Glass Loss Polyme Dirty Broken Disc Insulator Pollution Fla	88.2 81.1 87.7 81.6 97.9 90.2	81.8 85.0 86.0 81.7 97.3 81.3	0.858 0.873 0.870 0.847 0.989 0.858

算法	预处理时间	推理时间	后处理时间
YOLOV8	0.5	5.7	1.5
本文算法	0.3	5.1	1.1