Research on Multi-type Defect Detection of Insulator for Transmission Line Based on GE-YOLO Algorithm

WANG Qiang; FENG Yuntao; ZHOU Chao

doi:10.13296/j.1001-1609.hva.2025.06.011

您当前的位置：

首页 >

文章列表页 >

Research on Multi-type Defect Detection of Insulator for Transmission Line Based on GE-YOLO Algorithm

Research & Analysis | 更新时间：2025-06-12

- Research on Multi-type Defect Detection of Insulator for Transmission Line Based on GE-YOLO Algorithm
- High Voltage Apparatus Vol. 61, Issue 6, Pages: 92-101(2025)
- 作者机构：
  
  1.三峡大学电气与新能源学院，湖北　宜昌　 443002
  2.智慧能源技术湖北省工程研究中心，湖北　宜昌　 443002
- 作者简介：
- 基金信息：
  
  National Natural Science Foundation of China(52077120)
- DOI：10.13296/j.1001-1609.hva.2025.06.011
  CLC：
- Received：25 November 2024，
  
  Revised：19 February 2025，
  
  Published：16 June 2025
- 稿件说明：
移动端阅览
WANG Qiang, FENG Yuntao, ZHOU Chao. Research on Multi-type Defect Detection of Insulator for Transmission Line Based on GE-YOLO Algorithm[J]. High voltage apparatus, 2025, 61(6): 92-101.
DOI：

WANG Qiang, FENG Yuntao, ZHOU Chao. Research on Multi-type Defect Detection of Insulator for Transmission Line Based on GE-YOLO Algorithm[J]. High voltage apparatus, 2025, 61(6): 92-101. DOI： 10.13296/j.1001-1609.hva.2025.06.011.

摘要

因线路绝缘子航拍图像目标尺寸较小、背景复杂等客观局限，致使传统检测算法及多数轻量网络难以准确识别目标缺陷。文中提出一种基于GE-YOLO的绝缘子多类型缺陷检测模型，该算法以YOLOv8s为基础、结合分组卷积和多尺度融合思想设计分组多尺度卷积（grouping multi-scale convolution，GMSC）并使用其对C2f进行改进，增强模型特征提取能力；在特征融合网络引入高效多尺度注意力机制（efficient multi-scale attention，EMA），提高模型目标定位能力；最后将损失函数替换为Wise-IOU以降低几何因素的过度惩罚，提高模型回归的性能。实验结果表明，优化后的网络模型计算复杂度基本不变，满足边缘算力需求，整体

mAP

提升5.5%，其中破损绝缘子

mAP

提升10.9%，文中算法能够更均衡地实现对绝缘子多类型缺陷的检测。

Abstract

Due to such objective limitations as small target size and complex background in aerial images of line insulators

traditional detection algorithms and most lightweight networks are difficult to accurately identify target defects. In this paper a multi-type defect detection model for insulators based on GE-YOLO is proposed. The algorithm is based on YOLOv8s and combines grouping convolution and multi-scale fusion ideas to design grouping multi scale convolution (GMSC)

which is used to improve C2f and enhance the model's feature extraction ability. The efficient multi-scale attention mechanism (EMA) is introduced into the feature fusion network to improve the model's target localization ability. Finally

the loss function is replaced into Wise-IOU to reduce the excessive penalty of geometric factors and improve the performance of model regression. The experimental results show that the computational complexity of the optimized network model remains basically unchanged

meeting the edge computing power requirements. The overall

mAP

is increased by 5.5%

in w

hich the

mAP

of damaged insulator is increased by 10.9%. The algorithm in this paper can achieve the detection of multiple types of defects of insulators more evenly.

关键词

Keywords

references

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Research on Multi Type Defect Detection of Transmission Line Insulators Based on GE-YOLO Algorithm

Insulator Defect Detection Algorithm for Multi-scale Feature Extraction Based on Improved YOLOv7

Multiple Defect Detection of Insulators Based on Improved YOLOv5 Algorithm

Research on Vibration Defect Detection Method of GIS Based on Singular Spectrum Analysis and Decision Tree

Research on Power Component Identification of Transmission Line and Defect Detection Algorithm Based on YOLOv6

Related Author

ZHOU Chao

FENG Yuntao

WANG Qiang

ZHAO Ruhai

LI Yongling

ZHANG Miao

LEI Jingfa

GUO Yihan

Related Institution

Sichuan Provincial Key Laboratory of Process Equipment and Control Engineering

Key Laboratory of Intelligent Manufacturing of Construction Machinery, Anhui Education Department

School of Mechanical and Electrical Engineering, Anhui Jianzhu University

College of Computer and Information, China Three Gorges University

Hubei Provincial Engineering Technology Research Center for Power Transmission Line, Three Gorges University

⁰