Research On Key Components Detection And Their Anomaly Identification Algorithm Of Transmission Lines Based On Aerial Images

Overhead transmission lines are the blood of my country’s power system,and the stability of its power transmission is related to the normal operation of the economy and society,and people’s daily life.However,because the transmission line is in the external natural environment for a long time,the power components on it are easily damaged,which affects the power transmission of the entire system.Therefore,it is very necessary to conduct regular inspections on transmission lines to detect damaged components and other safety hazards in time.However,the traditional manual inspection method has been unable to meet the current wide coverage of transmission lines.With the popularization of UAV technology and the development of machine vision technology,the intelligent transmission line inspection method using UAV equipped with image acquisition equipment has become a new development direction in power inspection.Among them,the research on automatic visual detection technology for aerial power line images is the core focus.In view of this,based on deep learning,this paper conducts in-depth research and analysis on key components detection and their abnormal state identification technology.Key components detection and their abnormal state identification technology refer to detecting the power components on transmission lines and identifying whether there are hidden dangers or abnormal conditions(such as parts rust,missing,etc.)that can lead to power transmission faults against the complex background of aerial images.Firstly,this paper expounds the national strategic background of the high-speed development of overhead transmission lines in my country.And then implemented into the significance of people’s livelihood and welfare of the construction of transmission lines in Hunan Province,pointing out the importance and necessity of research the detection technology of key components of transmission lines.Afterward,the research status of domestic and foreign transmission line inspection and the development of object detection technology is introduced,and the advantages and disadvantages of several representative object detection methods are mainly analyzed,which lays a theoretical foundation for the detection algorithms in the follow-up research of this article.The main contents of this article are as follows:(1)Aiming at the actual needs of power transmission line inspection,we built a transmission line image capture hardware system with UAV as the main body.We also conduct an in-depth analysis of the characteristics of the collected aerial images of transmission lines and the power components on them and clarify the practical problems and challenges encountered in the inspection of transmission lines.(2)Aiming at the scale differences between different power components and the influence of complex background information in aerial images on the accuracy of object detection,a detection algorithm for key components of power transmission lines embedded with an attention mechanism is proposed to complete the high-precision detection of 8 different power components on transmission lines.(3)Aiming at the detection of abnormal conditions of key components on transmission lines,we research a multi-task detection algorithm based on region-guided detection that detects transmission line components while considering the identification of their rust levels and self-damage status(e.g.,dislodged,missing,self-explosion,etc.),and realizes the multi-visual word retrieval of power components.(4)Aiming at the problem that the feature details of small object components on power lines(referred to as bolts in this paper)are easily lost when deep neural networks extract features from aerial images,a detection algorithm for power line bolts and their defects based on multi-scale feature fusion is studied,and solves the problem of insufficient detection of small object components in power line inspection.To sum up,several detection algorithms proposed in this paper comprehensively consider power components with different distribution conditions and their unique characteristics on the transmission line,and put forward targeted detection and identification scheme.The experimental results show that the research content of the paper can well meet the actual needs of transmission line inspection.