Detection Of Insulator Contamination Based On Information Fusion Of Hyperspectral And Infrared Images

With the aggravation of air pollution,flashover accidents occur frequently,which seriously threatens the safety and stable operation of the power system.Pollution flashover prevention is a problem that needs to be solved in the construction of smart grids,and it is also the focus of researchers.Accurately assessing the contamination degree of the external insulation is necessary to arrange the cleaning cycle reasonably,and provide theoretical guidance for the prevention of pollution flashover.Traditional pollution detection methods have many shortcomings.Based on information fusion,this paper explores the feasibility of hyperspectral imaging technology and infrared imaging technology to establish a evaluation model for the contamination of insulators,which promotes new progress in the evaluation of external insulation contamination.In this paper,insulators of different pollution levels were prepared by artificial smearing,and the withstand voltage test was carried out.The hyperspectral and infrared images of the polluted insulators were collected.Using image processing technology,the collected images are corrected,denoised,segmented,which effectively reducing the interference of factors such as environment and instrument noise.By feature extraction,the hyperspectral and infrared image features that can effectively detect the pollution degree of insulators.Shuffled Frog Leaping Logrithm can extract the key information of the hyperspectral image,and convolutional neural network algorithm can extract the key information of the infrared image.Based on the information extracted by Shuffled Frog Leaping Logrithm and convolutional neural network algorithm,two detection models based on input layer fusion and feature layer fusion are established respectively.The accuracy of the two different fusion models is higher than the detection model based on the single source image feature before data fusion,indicating that information fusion can increase accuracy of the model.By comparing two different fusion models,it is found that the accuracy rate of the model based on the input layer fusion is 90%,and the accuracy rate of the pollution degree detection model based on the feature layer fusion is 93%.The model of feature layer fusion has higher accuracy,and the performance scores of performance indicators such as accuracy rate and recall rate are better than models based on input layer fusion.Based on hyperspectral image and infrared image information,combined with multisource image processing technology,this paper describes the pollution degree of insulators from different angles,and establishes a detection model of insulator pollution degree based on the fusion of hyperspectral and infrared image information,which breaks the bottleneck of traditional detection methods.It can provide theoretical guidance for unmanned aerial vehicle inspection of external insulated transmission lines.