Insulator Image Segmentation And Self-detonation Fault Detection Based On Machine Vision Technology

Insulators are an important part of the transmission line,and play a supporting and insulating role on the line.Due to their long-term exposure to the field,the influence of strong electric field,temperature and humidity,and defects in production,insulators have a high frequency of failure,so regular inspection is required every year.In this thesis,computer vision technology is applied to the detection of insulators and insulator images are studied.The details are as follows.(1)Firstly,the insulator image is preprocessed.After gray optimization,the image is denoised using four algorithms.Finally,the image is enhanced by comparing the denoising method with PSNR and SSIM parameters.(2)Two methods of edge image segmentation and threshold image segmentation are introduced,in which the two methods are analyzed and processed by different operators respectively,and the image is tested,and the suitable segmentation method is selected by comparing the advantages and disadvantages of several methods.(3)Study on the extraction of insulator images.The image morphology algorithm is introduced,and the insulators are extracted by color space transformation and color difference method respectively due to the different types of insulators.In view of the poor effect of the previous method,the color difference method is improved,and the selection of improved parameters makes the effect greatly improved.(4)YOLOv4 algorithm is used to classify and identify the insulator fault images.After training and improvement of the algorithm and verification of the on-site insulator images,it is found that the identification of the insulator self-explosion fault has obvious effect.In this thesis,the machine vision technology is used for experimental analysis of insulator images,and the segmentation and state detection of insulators are studied,which provides methods and ideas for the correlation of smart power grid.