| The environment in which insulator strings are applied is complex and variable.Insulator strings are affected by strong electric field and mechanical load,as well as the limited structure of the insulators itself,and the foraging of the bird.Therefore,insulator strings are extremely vulnerable to breakage especially in the umbrella skirt part.If these damages are not discovered in time,they will develop into more serious damages and cause large accidents.With the development of unmanned aerial vehicle technology,it is of great significance to identify insulator damage intelligently by image method.This paper mainly aims at image recognition of the breakage of composite insulators umbrella-skirt in overhead transmission line.In this paper,the following aspects were studied:First,the preprocessing process of insulator strings image was introduced.The best filtering method is selected by comparing the three denoising methods.The traditional method of maximum between-cluster variance was improved to enhance the effect of image segmentation of insulator strings.Secondly,the image of damaged umbrella-skirt of insulator strings was recognized.The equal-voltage ring matching method based on gray value was adopted to obtain the number of insulator strings in images.Based on the symmetries of peak-valley,the damage of outer contour of insulator s was identified.Scanning method was used to identify the damages of insulator strings in service.Finally,the cracks in umbrella skirt insulators were identified.Five kinds of characteristic values of cracks were given according to the characteristics.The cracks of insulators were divided into five types: transverse cracks,vertical cracks,mesh cracks,block cracks and none-crack.The types of cracks were identified based on BP neural network.The performance of different improved algorithms based on BP neural networks is compared.Through the crack type identification experiment,the correct rate of the model was verified. |
PDF Full Text Request