Abstract:As an onsite condition monitoring method, an infrared inspection can help to discover and analyze abnormal temperature increases in power equipment. For improving the efficiency of the onsite diagnosis of insulators in substations, this article proposes an automatic diagnosis method using instance segmentation and temperature analysis of infrared insulator images. For developing this method, thousands of infrared images from field inspection databases were collected to establish an annotated data set of insulator images. With the aid of the Mask R-convolutional neural network (CNN), it was possible to extract multiple insulators automatically in the infrared images. Transfer learning, as well as the dynamic learning rate algorithm, were then employed to realize the training process of Mask R-CNN with the annotated image data set. The result of the testing experiment showed that the mean Average Precision (mAP) of the model is 0.77, and the frame per second (FPS) is 5.07, which indicated great identification accuracy and computing speed of the proposed model. Next, function fitting was realized to extract the temperature distribution of each insulator. Finally, to evaluate the condition of each insulator, rules, which are based on the related standards, were established using machine language. This is the first time that the machine could independently realize fault analysis of multiple insulators in the infrared images, which is a great attempt to adapt the development of the Internet of Things and the tendency of predictive maintenance. Moreover, because of the universality of the model algorithm used, automatic infrared fault diagnosis for other power equipment could also be performed in a similar manner, which has significant potential applicability in the area of power equipment diagnosis.

Box-Point Detector: A Diagnosis Method for Insulator Faults in Power Lines Using Aerial Images and Convolutional Neural Networks

Detection of Power Line Insulator Defects Using Aerial Images Analyzed With Convolutional Neural Networks

A Multi-feature Fusion-based Deep Learning for Insulator Image Identification and Fault Detection

Defect Insulator Detection Method Based on Deep Learning

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

Arbitrary-Oriented Detection of Insulators in Thermal Imagery via Rotation Region Network

Deep Learning Approaches on Defect Detection in High Resolution Aerial Images of Insulators

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

A Convolutional Neural Network-Based Defect Recognition Method for Power Insulator

A deep learning approach for insulator instance segmentation and defect detection

Railway insulator defect detection with deep convolutional neural networks

Defect Recognition of Insulators on Catenary Via Multi-Oriented Detection and Deep Metric Learning

An method for power lines insulator defect detection with attention feedback and double spatial pyramid

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

Image Detection of Insulator Defects Based on Morphological Processing and Deep Learning

Insulator Defects Detection for Aerial Photography of the Power Grid Using You Only Look Once Algorithm

Power line insulator defect detection using CNN with dense connectivity and efficient attention mechanism

Defect Identification Detection Research for Insulator of Transmission Lines Based on Deep Learning

Accurate Glass Insulators Defect Detection in Power Transmission Grids Using Aerial Image Augmentation

The method of insulator recognition based on deep learning