Research On Discrimination Method Of Insulator Pollution Grade Based On Visible Image Recognition

Growing natural environment pollution directly reveal the insulator surface in atmosphere in accumulated filth faster,degree aggravating,Moreover,increasingly promote transmission and improve level of line voltage and speed up the expansion of the electrified railway catenary,the possibility of insulator pollution flashover accident was greatly increased.Therefore,in order to reduce insulator pollution flashover accident,it is necessary to timely detect the pollution status of insulator surface.This paper combines digital image processing technology with artificial intelligence algorithm,takes visible light image of insulator obtained by artificial pollution experiment of insulator as the research object,and is committed to building a model of insulator pollution level discrimination based on visible light image recognition.In this paper,the corresponding relation between visible image of insulator and pollution state of insulator is analyzed firstly.Artificial pollution experiment is carried out on insulator in an ideal laboratory environment,and a large number of visible image of insulator are acquired by using a camera with superior performance.Secondly,in view of the continuous improvement of the performance of visible light imaging equipment,the processing process of insulator image has been improved,that is,as long as the image is converted,decomposed,segmented and denoised,the target features of the image can be highlighted.Third in order to select the best feature of the visible light image information in order to improve the determination accuracy of insulator pollution level,the characteristics of multiple images was calculated on the basis of pollution level and the statistics and analysis,and their classification by fisher criterion function calculation effect value to determine the characteristics of selected,obtained with pollution level training set and testing set data labels.Finally,based on least squares support vector machine(SVM),design a few intelligent multi-valued classification models,and for this a few classification model input respectively test the training set data and the parameter optimization,and then input test set data classification performance testrespectively,finally,the output of the insulator pollution level of classification accuracy are analyzed in comparison.Selected from the gallery is 0,Ⅰ,Ⅱ,Ⅲ,Ⅳ all the pollution level of visible light image 300 pieces and the average is divided into two groups,a group of images for training set,another set of images for the test set.The intelligent multi-valued classifier of least squares support vector machine based on genetic algorithm optimization,the intelligent multi-valued classifier of least squares support vector machine based on particle swarm optimization and the intelligent multi-valued classifier of least squares support vector machine based on beetle antennae search algorithm optimization are designed,and the training set data and the test set data were used to study and train them successively and to test the classification of pollution level.The classification accuracy of the three classifiers reached 93.33%,94.67% and 96.00%,respectively.Therefore,the research method in this paper can realize the detection of insulator pollution degree in the pollution deposition stage,and the detection effect is good.