Identification Of Insulator Damage Based On Image Processing

In recent years,electrified railway has been applied more and more widely,The working state of the insulator in the catenary is very important to the normal operation of the electrified railway.Insulators have been exposed to the field for a long time,so there are safety accidents such as self-explosion,damage and flashing,which lead to safety accidents such as self-explosion,cracks and pollution of insulators.The traditional manual inspection method is a waste of manpower and may not achieve good results.In this paper,the image processing technology is used to identify the insulator in the complex background,and then identify the defective insulator.Firstly,the photos of the insulators in the complex background of the laboratory are used as experimental pictures.Because of the noise and uneven light of the captured images,the experimental pictures are preprocessed,mainly including image graying and denoising filtering,to make the image features more visible.Secondly,according to the limitation of Harris corner matching algorithm,an improved algorithm combining SURF description operator and corner matching is proposed to extract and match image feature points.Through experimental comparison,it is found that the corner matching algorithm combined with the SURF description operator can well identify insulators in complex backgrounds.At the same time,this paper also proposes an insulator extraction algorithm based on region of interest recognition.It is found that this method can avoid the existence of more false targets and over-segmentation in threshold segmentation.Finally,in terms of insulator classification,BP neural network is easy to fall into local minimum value and many iterations in the learning process.This used BP neural network optimized by particle swarm optimization.Through experimental analysis,PSO-BP neural network are superior to BP algorithm in operation speed and recognition accuracy.To a certain extent,it solves the problem that the manual detection of broken insulators is inefficient and easy to miss detection.