Application Research On Detection And Extraction Of Aerial Insulator Images Based On Convolutional Neural Network

Transmission lines are the main hub of power transmission in China.Insulators,as an important part of them,can not only provide mechanical support for power lines,but also prevent current from being grounded.However,due to the structural characteristics of the insulator and the external environment and other factors,the self-explosion,flashover and other faults are easily caused,and thus the transmission line cannot operate normally.Therefore,it is very important to inspect the state of the insulator on time.In the process of insulator detection and identification,the selective search method is used to propose and segment the image generation target,then designs the structure of the convolutional neural network for the insulator identification,and combines this method with the HOG feature to combine the SVM classifier and the PCA.A comparison was made with the identification method of BP neural network.The representation of HOG combined with SVM classifier is a traditional image processing method.PCA combined with BP neural network represents a traditional full-connection neural network method.In the process of detecting the lack of self-explosion of the insulator,the reconstruction of the insulator information is achieved by deconvolution operation according to the parameters of the trained convolutional neural network,and the reconstructed saliency map is super-pixel segmented,and then the mathematical model of the insulator string is established.Finally,the precise location of the insulator's self-explosion is lost.The experimental results show that the method of using convolutional neural network is the best among the algorithms of insulator detection and recognition.This method can basically avoid the influence of shooting distance,angle and other conditions,and has good recognition ability for complex background.Therefore,this paper finally proposes an insulator self-explosion loss detection algorithm based on convolutional neural network identification of insulators,and finally achieves an accurate positioning of the insulator's self-explosion missing position.