Research On UAV Visual Localization And Flight Monitoring Methods For Road Inspection Applications

In recent years,with the continuous development and progress of the intelligent level of unmanned aerial vehicle(UAV),the inspection technology based on UAV has begun to be widely used in the fields of electric power inspection,crop monitoring,aerial surveying and mapping,forest fire monitoring,pipeline inspection,and so on.However,in terms of highway maintenance,UAV mostly rely on operator controlling for road inspection.With a low degree of automation,this way is difficult to meet the needs of large-scale highway maintenance.UAV signals that rely on GPS for navigation are susceptible to interference from the surrounding environment,which can easily cause UAV to deviate from the road and affect the quality of road inspections.Combining UAV inspection technology with computer vision and deep learning technology can fully exploit the advantages of UAV such as strong mobility,low power consumption,and high degree of automation.Based on this,the thesis studies the visual positioning and flight control methods of UAV for road inspection.Firstly,the road identification and UAV lateral positioning are completed using the road images captured by the UAV.And then the lateral offset is adjusted through the UAV flight control algorithm to maintain the autonomous flight of the UAV along the center of the road.The main work of the thesis is as follows:(1)Proposed a lateral positioning method for UAV along the road based on road identification and center of gravity calculation.The Deep Labv3+ network model is selected to identify the road areas in different types of road images captured by UAV.The contour detection and polygon fitting methods are used to calculate the center of gravity of the road areas.Then the camera calibration and coordinate change methods are used to locate the road areas.Finally,the accuracy of the road positioning is verified through experiments.(2)A control algorithm for UAV flying along the road is designed,with visual positioning as the main approach and GPS positioning as the auxiliary approach..The lateral control uses PID algorithm to adjust the lateral offset.The combination of azimuth,heading angle,and road inclination controls the rotation angle of UAV.The intersection uses waypoint flight control to select the correct heading.Software in the Loop(SITL)simulation experiments and actual road flight experiments show that the control algorithm proposed in this thesis can achieve autonomous flight of UAV along the center of the road.(3)A remote monitoring and control software for UAV based on the Windows platform has been developed.The monitoring and control software is developed based on Py Qt application.The specific functions include remote monitoring of road images and road recognition,real-time display of drone flight status data,remote basic control of drones,and extraction of flight waypoint data after path planning based on the Gaode map.When abnormal situations occur,road maintenance personnel can remotely control drones to fly to the center of the road surface through software.The software functional testing results indicate that the remote monitoring software developed in this thesis can ensure the security of inspection tasks.