Research On A Fault Detection Method Of Transmission Line Component Based On The Aerial Images

In recent years,the scale of investment in China’s power grid has been expanding,resulting in a more complex power network structure.While the smooth transmission lines have an inseparable relationship with the normal operations of various components.At present,the fault detection of transmission line components in China is mainly to design the corresponding feature extractor for the specific component image collected by aerial photography,and then classify and judge.In view of the diversity of components and the complexity of background in aerial image acquisition,the limitations of using only the unique feature extractor are too great and it is also difficult to guarantee the robustness of complex background image.Meanwhile,due to the strict regulations of low-altitude areas in China,the personnel of aerial inspection cannot freely patrol practice.Therefore,this paper puts forward a fault detection method of transmission line components based on the aerial images to solve the above problems.This paper focuses on building a network model for fault identification and detection of transmission lines to carry out intelligent detection of fault component.At the same time,as for the control of low-altitude in China,this paper proposes the inspection training of transmission line based on VR(Virtual Reality)technology,which can effectively alleviated the limitation of personnel’s inspection inability to carry out inspection.Firstly,based on the transmission line component images collected by aerial photography,the wire-18 DS is established by the 3D modeling and rending under different backgrounds.Secondly,based on the regional proposals network algorithm,a network model of transmission line components identification and fault detection is built,and a fault identification network with strong generalization is obtained by training the parameters on the self-built database.Then,the trained model is loaded into Qt to automatically detect and classify the aerial image,generating the corresponding fault report.Meanwhile,it also judges the tower rod of the detected fault part through the longitude and latitude information.Finally,virtual reality technology is used to build 3D models of real terrain,transmission towers,components and UAV(Unmanned Aerial Vehicle),and design the interaction between HTC VIVE handle and the UAV in virtual scenes,so as to simulate and control the UAVs’ inspection of transmission lines.The experiment results indicate that the fault detection network model for transmission line components designed in this paper can simultaneously detect eighteen common components such as insulator,grading ring and counterweight,and also achieve an average recognition rate of 91.1% in the self-built Wire-18 database.At the same time,transmission line inspection based on VR technology also realizes the direct human-computer interaction between the external handle and the virtual scene.Users can immerse themselves in the operation steps of aerial photography inspectors on the UAV.