Design And Implementation Of Driving Assessment System For Power Line Patrol UAV

My country’s power industry has long relied on manual inspections,but with the complexity of the national power grid,manual inspections are obviously not suitable for the development of the power industry.As China Southern Power Grid and State Grid vigorously promote drone inspections,a production line operation and maintenance model with "machine inspection as the mainstay and manual inspection as the supplement" is being formed.Since the safety and efficiency provided by drone patrols cannot be achieved by manual patrols,drone patrols have been put on the agenda to completely replace manual patrols.However,due to the complex working environment of the patrol drone and high requirements for operators,the State Grid needs to conduct patrol training for pilots,and need to pass a special assessment system to judge the flight trajectory deviation..In this paper,a line-following UAV driving assessment system is designed,through the non-contact position detection of the line-following UAV by the visible light camera,and the trajectory error is compared,to realize the evaluation of the piloting ability of the pilot UAV.The system combines technologies such as camera calibration,rendezvous measurement and target detection.Through theoretical analysis,simulation calculation and experimental verification,it conducts in-depth research on the detection of UAV targets and spatial position measurement.The main research work of this paper includes:1.Aiming at the problem that the current camera calibration method has low calibration accuracy for wide-angle lenses with large distortion,this paper uses the biharmonic spline interpolation method to design a camera calibration scheme with high accuracy and suitable for drone flight assessment scenarios.In this method,a highprecision RTK positioning module is mounted on the drone,and several sample points that can represent the whole area are collected within the field of view of the camera lens,and then the information of the full-size pixel points is fitted by interpolation.2.In view of the low detection rate of small targets,slow detection speed and more susceptibility to noise interference by current target detection algorithms,this paper uses deep learning to detect UAV targets using convolutional neural networks.Build a UAV data set around actual application scenarios.The data set includes multi-scale,multiposture actual application scene pictures,network pictures and open source data set pictures,and the detection of UAVs is improved through data enhancement and fusion of background differences.3.Combining the above-mentioned camera calibration,rendezvous measurement and target detection methods,finally completes a line-following drone driving assessment system that can sample the flight trajectory of the drone,and then evaluate the deviation between the flight trajectory and the predetermined trajectory.The line inspection drone driving assessment system designed in this paper has good flexibility and stability,and at the same time has better drone detection effects and positioning accuracy.Compared with the existing State Grid’s assessment system,it can measure flight better.Trajectory deviation has high practical application value.