Porcelain Insulator Contamination Grades Detection Based On Infrared-Thermal-Image Processing

Along with the extensive generalize of high voltage or extra-high voltage power transmission and the increasement of transmission circuit voltage grade,some pollution flashover accident of power grid occurred frequently.Therefore,efficient and accurate access to insulator contamination grades is essential to ensure the safe operation of the grid.At present,the domestic and overseas detection methods to measure insulator contamination grades mainly includes ESDD method,leakage current measurement,and spectral method and so on.Although the above method has a certain feasibility,but they basically react to the single insulator to processing,and ignore the heat generation characteristic of polluted insulator chain and change in appearance of exothermic region.Hence,the current detection methods usually has complex operation and high misjudgment rate.In order to address these issues,in accordance with the heat characteristic of contaminated insulators,a method based on color image processing is put forward.In this paper,using the methods based on color image processing,the pollution grade of insulators is judged by processing these infrared image of contaminated insulator.(1)Using two-step method,perform the de-noising process on the infrared image of contaminated insulator.Firstly,perform median filter to remove impulse noise;secondly,perform wavelet diffusion method to remove Gaussian noise;(2)In the light of the color characteristics of insulator infrared thermal image,improved Otsu segmentation algorithm that combine the traditional OSTU algorithm with wavelet transform and R component chart composed by the R component of RGB color image are used to obtain a binary image.(3)The resulting binary image is amended twice,and the correction procedure is based on mathematical morphology.(4)Using the methods in statistics to obtain the five characteristic components that belongs to the R channel of the dirt retention area,which including perimeter-to-area ratio,minimum and maximum,extrema,standard deviation and the minimum axial ratio of ellipse.Then,training the BP neural network and establish a pollution level detection model for color image.In this paper,when the administration power supply is running in Qitaihe,we conduct the infrade image to the insulator.Wherein the shooting condition selected relative humidity of 78%~91%.Due to the influence of temperature on the thermal infrared imager is not great,so the temperature can be controlled at room temperature,between 10 ? ~ 30 ?.After shooting an infrared thermal image each time,the pollution level is recorded in accordance with the GB /T5582-4993.Respectively,choose 500 groups of XP-70 porcelain insulator infrared thermal image which own 0,?,?,?,? pollution level as the sample,repeat the experiment for the system,and the final grading screening accuracy of 91%.Experimental result demonstrates,combine the appearance of the polluted insulators with the characteristics of the infrared thermal image is contribute to improving precision of grading screening,and it laid some foundation of further study on the pollution levels division of the insulators under complex environment.