Research On Image Feature Classification And Ice Detection Algorithm Of Aerial Transmission Line Based On DCNN

Transmission lines are an important channel for national power transmission,and wires and insulators are an important part of it.Since most of the transmission line equipment has built outdoors,it needs to withstand various external environments and various natural disasters.Among them,due to the intrusion of ice and snow,the ice coating value of the transmission line exceeds the set critical value,resulting in wire breakage,tower collapse,insulator flashover and other accidents,so that the transmission line often fails to operate normally.Therefore,the State Grid Corporation needs to spend many human resources to patrol the transmission line,remove the bird's nest on the tower,and remove the ice on the wires and insulators.China has a vast territory,complex terrain,large climate differences,and a wide distribution of transmission lines,which has caused great trouble to the inspection work of the staff.In order to assist the inspection work of power grid staff,this paper presents an algorithm for feature recognition and classification of UAV aerial transmission line images based on deep convolution neural network(DCNN is replaced later)by image processing technology,and calculates the icing image in detail.Firstly,a transmission line image data set will be established,and the data set is divided into pre-training set and test set.Secondly,the images in the pre-training set are labeled and classified manually by using the image labeling tool built by QT5 and Python.Secondly,the images in the pre-training set are input into the depth convolution neural network,and combined with the improved SVM multi-classifier,a transmission line image classification model is trained and generated.Then,the test image is input into the model to get the category of the image;the image is input into a judgment model and skipped if the image is normal;if the image is an icecovered image or an image containing a bird's nest,the image is output;finally,the output image is further processed to obtain the actual coordinate information of these images.Used to guide staff to do a good job of obstacle removal.The experimental results show that the image classification algorithm based on deep learning for UAV aerial transmission line can classify the transmission line image well.The ice thickness calculate method can also accurately calculate the ice thickness of transmission line image.The calculated error has some optimization compared to the current work.Therefore,the proposed algorithm has certain reference value for the classification of transmission line features and the calculation of ice thickness of good insulator images.