Research Of The Detection Method Of The Faulty Insulator Based On Infrared Image

As important parts of overhead transmission lines, insulators play an indispensable role in safeguarding the electrical power system. Therefore, it is significant of social effects and economic values to detect and replace faulty insulators in time. However, traditional detection methods of the faulty insulator have many shortcomings, such as great workload, high cost, non-charged detection and low security, so they cannot satisfy increasing requirements. Fortunately, as the growing popularity of infrared thermograph technology, detection methods for faulty insulators based on infrared image are becoming new research hotspot.This paper researches the existing detection methods of faulty insulators systematically and deeply, and proposes a new detection algorithm based on infrared image, whose feasibility and veracity are verified by experiments. In general, the detection algorithm is mainly divided into three steps including preprocessing of the insulator infrared images, extraction of insulator areas in complex background as well as classification of the faulty insulators based on BP_Adaboost.As for the preprocessing of the insulator infrared images, the paper has researched the common frequently-used preprocessing methods, and proposed an improved local HE algorithm involving the characteristics of the insulator infrared images, as is used to improve visual effect of the infrared images.As for the extraction of insulator region, this paper has researched the existing methods, and analyzed these methods were not applicable to complex background. To address this problem, an algorithm to extract the insulator region under complex background is put forward, as is a combination of multifarious mathematical tools, such as GA, mathematical morphology, Hough transform, ellipse detection, etc. Moreover, the experimental results show that this algorithm is of high accuracy.As for the classification of the faulty insulators, this paper has researched BP neural network as well as Adaboost model, and constructed a classification model based on BP_Adaboost, using the temperature distribution data of the insulator region to train and test the model. The experimental results show that the proposed model is of higher recognition accuracy than traditional methods, too.Finally, a faulty insulator detection system has been implemented based on MATLAB, whose functional modules were designed independent subsystem in order to enhance their scalability, universality and low coupling. What's more, the detection system is packaged visually based on MATLAB GUI, so it could be debugged and be used to detect faulty insulators more intuitively.