Pollution state detection of insulators based on multisource imaging and information fusion

Abstract

The insulators' pollution flashover will cause huge economic losses. But at present, there is no non-contact method with high accuracy to detect the pollution state of insulators. This paper aims to realize the non-contact online detection of the pollution state on the surface of the insulators, by researching on multi-source imaging methods, including visible imaging, infrared imaging and ultraviolet imaging. The insulators were polluted according to the IEC standard, and the visible images of the polluted insulators were obtained to find the features of the images. After that, the polluted insulators were tested with their working voltage. At the same time, both the infrared images and ultraviolet images were shot, in order to get the features of the images and the relationship between the features and the pollution state of insulators. Finally, a BP neural network was set up, fused by the three kinds of imaging detection methods, and an accuracy test was conducted. Before the fusion, the accuracy of every one of the three imaging detecting method was no more than 85%. However, after the fusion, the accuracy of multi- source imaging detection rose to 90% and the incorrect detection disappeared.