Power Tower Identification And Measurement Based On Drone Patrol Inspection

At present,the measurement of unknown towers in the power grid is mainly carried out by UAV equipped with lidar to model the power pole tower in three dimensions,and then obtain the parameters of the unknown tower,but lidar is expensive,according to the method of this article,there is no need to carry lidar,which saves a lot of money for the purchase of hardware.In order to solve the above problems,this thesis proposes a UAV fully autonomous power pole tower recognition and measurement method based on monocular camera shooting,which mainly includes: target detection for unknown towers,judgment of tower types,and calculation and solution of tower parameters,as follows:Firstly,different power tower structures correspond to different key point selection strategies,and the four tower types and corresponding key point selection principles of 220 k V and above voltage levels mainly studied in this thesis are analyzed.Based on the research on the production,image management,image annotation and image enhancement technology of the tower dataset,the UAV line tower inspection image database was constructed.Aiming at the problem of insufficient dataset richness caused by too small dataset sample size and uneven distribution of data categories,random changes in brightness,contrast and color are used to increase the image sample size,and finally form an image sample set with uniform categories and moderate quantity suitable for this research work,so that it can be learned under different conditions,so as to be more versatile.Secondly,a power pole tower identification and measurement scheme based on UAV inspection is proposed.The scheme includes tower target detection,type determination,and critical point detection.Firstly,the target recognition algorithm based on YOLOv5 is used to identify the unknown tower,then the tower image classification algorithm based on MobileNetV3 is used to determine the type of the power tower,and finally the key point detection algorithm based on HRNet is used to detect the key point of each type of tower and output the key point information,which lays a foundation for the next step of using the key point information to solve the tower parameters.Experiments show that compared with other classical algorithms,this method has significant improvements in accuracy,recall,loss and other aspects.Finally,the corresponding parameters of four types of tower are measured.Based on the photos taken by the UAV after displacement in accordance with certain rules,combined with the pixel coordinates,geographical position and spatial position of the UAV,the corresponding coordinate transformation of the same pixel in the pictures was carried out,and the parameters of the tower were calculated.In addition,this thesis also analyzes the error caused by calculation and puts forward the solution.Through the analysis and calculation of four different power tower examples,the maximum relative error between the measured value and the real value of the tower is less than 5%,which ensures the feasibility of the proposed algorithm.It provides the basis for the parametric modeling of power tower in the future.