Research On Judgment Method Of Hydrophobicity Grade Of Composite Insulator Based On Intelligent Image Recognition Technology

Composite insulators are widely used in anti-pollution flashover applications of transmission lines of various voltage levels due to their excellent pollution flashover resistance,which improves the anti-pollution flashover ability of the power grid.However,the hydrophobicity of composite insulators is not constant.With the gradual aging of the silicone rubber umbrella group of composite insulators,its hydrophobicity is reduced,which increases the probability of pollution flashover accidents,and seriously threatens the safety and stability of our country power system run.On the basis of summarizing and analyzing the previous research on the hydrophobicity level detection of composite insulators,this paper conducts research on image denoising,image enhancement,image segmentation,intelligent classification and determination related issues involved in the water-repellency level determination of composite insulators based on intelligent image processing technology.First,in order to improve the contrast of the sample image,this article will denoise and enhance the sample water spray image.Under the premise that the samples are processed into the same specification,wavelet denoising and non-average local filtering are performed on the image.This method effectively separates the detail information and the noise information,and solves the problem of uneven illumination of water droplets or water film images.By adopting the method of adaptive histogram equalization,the local contrast of the image is improved and more image details are obtained,and the purpose of revealing the details of the shadow area is also achieved.Secondly,the improved particle swarm optimization algorithm and the twodimensional Ostu threshold method are combined to increase the calculation speed of the threshold optimization,thereby improving the efficiency of image segmentation.The improved Canny edge detection algorithm is used to accurately distinguish the target pixel from the background pixel with the best threshold obtained,and the edge of the water droplet is completely extracted.The method of mathematical morphology is introduced to completely label the edge of the image,which solves the problem of broken lines at the edge of the image.Finally,use the sample data set to train the constructed hydrophobicity level recognition model,and then use the test data set to verify,and use the improved VGG-16 deep convolutional neural network to complete the hydrophobicity level determination of composite insulators.The test data shows that the recognition rate of hydrophobicity level has reached over 93%.The deep learning method can automatically extract features from the sample data set.After comparing with the training results of the traditional convolutional neural network,a high recognition rate can also be obtained for small sample data sets,which greatly improves the hydrophobicity level of composite insulators.Judgment efficiency.