Research On Automatic Crack Detection Technologies Of Porcelain Insulator Based On Electricity Inspection UAVs

Insulators are important facilities in transmission lines.Due to the exposed outdoor environment,it is vulnerable to many unfavorable environmental factors,resulting in various types of faults.However,due to material differences,porcelain insulators are more prone to pollution flashover,fracture and other faults than other types of insulators,so fault diagnosis of porcelain insulators is an important link in power line inspection.With the maturing of UAVs technology,more and more people are welcome to adopt UAVs to inspect power lines.In this paper,unmanned power line inspection as the background,the crack on the surface of porcelain insulators as the research object,based on the inspection of the UAVs electric insulator crack detection technology.These technologies are generally divided into two modules,the first is to find the target of the porcelain insulators from the images taken by the UAVs,the object detection technology;the second is to analyze and extract the crack information from the close-up images taken by the UAVs,That is,crack detection technology.Object detection technology is achieved by constructing a convolutional neural network model based on region proposals.This paper proposes a Faster R-CNN model that is tailored to specific mission requirements.Through the pretreatment of the image set and the adjustment of the network structure and parameters,a TensorFlow-based deep learning architecture training model is adopted to realize the recognition and localization of insulator objects in the images taken by the UAVs.As for the crack detection of porcelain insulators,two schemes are proposed in this paper.The two programs have in common the idea of using edge detection as the basic idea of crack detection.The first solution is based on a variety of traditional image processing algorithms.Firstly,the original image is preprocessed.Then,the edge detection operator is used to calculate the edge gradient in the image.After that,the image filtered by the edge detection operator is transformed into a gradient image.Finally,the pixel where the crack is located is filtered out by threshold processing.The second option is based on a more intelligent ant colony algorithm.Firstly,the basic principle and existing problems of the traditional ant colony algorithm are introduced.Secondly,the ant colony algorithm is applied to the image edge detection by constructing the model.Among them,the foraging rules of ants,rules of pheromone release and some corresponding parameters have been improved,and finally the crack character of porcelain insulators has been extracted.