Research On Integrated Control Of Hybrid UAV For Power Line Corridor Detection

Due to the year-round exposure of transmission lines,accidents such as tripping and tower collapse caused by various construction and violations have been increasing year by year,seriously threatening the safe and stable operation of transmission lines,Therefore,it is more and more important to carry out line inspection work reasonably and effectively.Nowadays,unmanned aerial vehicles(UAVs)play an increasingly important role in the transmission line inspection,which mainly use the traditional four-rotor and fixed-wing UAVs.In order to effectively solve the problem of the short life of the traditional quadrotor drone and the strict requirements of the ordinary fixed wings on the landing site,this paper builds a new type of fixed-wing UVA platform that can take off and land vertically,which is called a compound unmanned Aircraft,and adopts an intergrated control scheme,which solve the problem that the previous compound UVA was not stable enough when switching flight modes.In addition,the drone will take a lot of photos when performing the inspection task.If it is only to manually identify and classify,the workload is very large,so this paper proposes an improved convolutional neural network for the classification of transmission channel inspection photos.The model can quickly classify the channel photos taken by the drone,so that the maintenance personnel can quickly deal with the corresponding defects.Based on the compound UAV,this paper studies the intergrated control scheme of the compound UVA platform and the classification method of transmission line patrol image.The first part of the paper explains the main components and flight principles of the compound UVA,and adopts an intergrated control scheme for the transitional flight phase,and the feasibility of the scheme is verified by simulation.Then,using the TALON mini fixed-wing carrier as a platform,a composite drone flight platform was assembled and built.According to the PID control principle for flight attitude adjustment in the PX4 flight control platform,the log records in actual flight were analyzed,and related parameters were adjusted,which improve the overall flight performance of the UVA.The second part carefully studies the convolutional neural network for image classification.Through specific experiments,we deeply understand the impact of each component of the network on the classification performance.Then compared with the existing network structure for image classification,VGG16 was selected as the classification model of transmission line patrol image,and a specific optimization method was proposed for the network according to the existing technology.In the third part,VGG16 was optimized according to the optimization method,and verifies and compares it through experiments,the best optimization scheme was obtained which achieves the desired effect of image classification.