Abstract:String insulators are interconnected disks that insulate the electrical conductors of high-voltage transmission lines from the supporting towers, withstanding both mechanical stress and electrical current. Regular cleaning is crucial for string insulators, as different types of pollutants can build up on the insulators, impairing their ability to prevent electrical arcing and short circuits. However, visual observation, the primary method for verifying cleaning needs, can be misleading. Developing robust algorithms for autonomous system-based inspection leveraging artificial intelligence presents a promising solution to address the limitations of visual inspection in preventive insulator maintenance. However, machine learning algorithms require considerable training data to learn from patterns, avoid overfitting, and account for outliers and data variation within the training dataset, especially considering the limited availability of images depicting dirty string insulators, as reported in the literature. In response to the limitations of existing training data for machine learning algorithms in insulator inspection, this work proposes developing a synthetic dataset for clean and dirty string insulators using software tools, such as SolidWorks, Inventor, and Unity 3D to model the towers, insulators, backgrounds, and various pollutant types. The clean string insulator dataset contains 47,286 synthetic images featuring diverse string insulator materials (glass, polymer, and porcelain) situated within various realistic backgrounds (mountain, forest, desert, city, river, and plantation). Similarly, the dirty insulator dataset comprises 14,424 distinct images, including clean insulators and simulated contaminants, such as soot, salt, and bird droppings. To test the proposed dataset, first, a classical neural semantic segmentation network was used to segment the clean string insulators dataset images, and an average dice coefficient of 0.95 was achieved by using only synthetic images. The same network, trained only with the synthetic dataset, was also tested using real-world images of string insulators, resulting in an average dice coefficient of 0.92. The synthetic dataset of dirty string insulators was used in a classical classification deep neural network, obtaining an average accuracy of 0.97 for real images of dirty insulators. This work demonstrates the potential of fully synthetic datasets for training machine learning models. Our results indicate that such models can achieve high accuracy in real-world applications, including accurately identifying insulator chains and classifying contaminants within real-world images.

DINS: A Diverse Insulator Dataset for Object Detection and Instance Segmentation

Catenary Insulator Defect Detection: A Dataset and an Unsupervised Baseline

FINet: An Insulator Dataset and Detection Benchmark Based on Synthetic Fog and Improved YOLOv5

Catenary insulator defect detection based on contour features and gray similarity matching

SODD-YOLOv8: An Insulator Defect Detection Algorithm based on Feature Enhancement and Variable Row Convolution

Object detection-based inspection of power line insulators: Incipient fault detection in the low data-regime

Missing-Sheds Granularity Estimation of Glass Insulators Using Deep Neural Networks Based on Optical Imaging

An Instance Segmentation Method for Insulator Defects Based on an Attention Mechanism and Feature Fusion Network

Efficient Cross-Modality Insulator Augmentation for Multi-Domain Insulator Defect Detection in UAV Images

Insulator defect detection algorithm based on adaptive feature fusion and lightweight YOLOv5s

Fault Detection Method of Glass Insulator Aerial Image Based on the Improved YOLOv5

Insulator Detection in Aerial Images for Transmission Line Inspection Using Single Shot Multibox Detector

Data Augmentation of Insulator Aerial Images for Intelligent Inspection of Electric Transmission Systems

Synthetic images datasets of clean and dirty string insulators used in high-voltage power lines

MG-MetaDETR: Marrying Muti-Scale and Grouping Strategy with DETR for Few-Shot Insulator Defect Detection

VISION Datasets: A Benchmark for Vision-based InduStrial InspectiON

Insulator Defect Recognition Based on Vision Big‐Model Transfer Learning and Stochastic Configuration Network

Detection of Glass Insulators Using Deep Neural Networks Based on Optical Imaging

A Casting Surface Dataset and Benchmark for Subtle and Confusable Defect Detection in Complex Contexts

High-resolution infrastructure defect detection dataset sourced by unmanned systems and validated with deep learning

Automatic Fault Diagnosis of Infrared Insulator Images Based on Image Instance Segmentation and Temperature Analysis