Study On The Method Of Fault Detection Based On Insulator Images

There comes a series of problems such as high consumption of human and material resources,poor safety,and low efficiency while using the traditional manual inspection method with the increasing of demand in electric power and expanding scale of transmission lines.Finding a more automated and intelligent power inspection method has become a hot research topic between the power-related workers and even the society in all circles.It is very significant to study the detection method of the insulator to provide a reference for the intelligent inspection of the transmission line.This paper mainly studies the ways to locate and defect faulty insulators with aerial insulator images and different fault detection methods.Finally the automatic detection of the "self-explosion" problem of the insulator will be realized.The work of this paper can be divided into three parts.At first,analyzing and processing the data.Combining the research process and the characteristics of aerial image data,the original images are pre-processed by a series of operations such as Duplicate removal,down-sampling,image enhancement,and noise removal to ensure the accuracy and validity of the basic data.Moreover,studying on the methods to defect fault based on image processing.The paper adapts an image processing method based on connected region,line fitting and morphology.Then designing the algorithm model and experiments to realize the insulator location and "self-explosion" detection.At last,researching on the “self-explosive” detection method based on deep learning.the cascading insulator detection model based on regional proposal is used to realize the location and detection of insulator according to the actual insulator data.Data augmentation is finished by affine transformation and background fusion to solve the uneven of positive and negative samples.The validity of the model was verified by multiple sets of comparison experiments.By preprocessing the aerial insulator images,the image processing and depth learning detection methods are designed and implemented to detect the self-explosion defects.The experimental results show that for the input insulator image,the insulator positioning and self-explosion area can be identified and showed in the image.The location of the defect helps the staff quickly find the broken insulator and replace it in time.