Infrared Image Fault Recognition Method for Disconnector Based on HOG Features

To realize the fault recognition of an infrared image of a disconnector, this study uses an improved SLIC algorithm to segment and mark the fault area of the disconnector based on color space conversion. As a result, image segmentation accuracy was significantly improved. Based on HOG feature extraction, the support vector machine algorithm is used to classify an infrared image of a disconnector and determine whether it works in the normal state. For the disconnector in the normal state, the relative temperature difference method is used to determine its fault state. The greater the relative temperature difference, the more serious the fault. The experiments demonstrate that optimal HOG characteristic parameters yield the maximum accuracy of the imaging equipment. The fault diagnosis of an infrared image can be used to determine the fault and defect degree of the disconnector and provide maintenance. The model used in this study exhibits good accuracy and reliability.