Autonomous Obstacle Avoidance,Tracking And Autonomous Positioning System Based On Multi-rotor UAV

In recent years,the research on unmanned aerial vehicle(UAV)has attached more and more attention owing to its characteristics such as low cost,good maneuverability,and flight stability.Together with the rapid developments of sensor technology,these issues on UAV's autonomous obstacle avoidance,tracking and positioning have become a heated topic in the academic field.Thus,aiming at the requirements of autonomous obstacle avoidance,target tracking and autonomous positioning,this dissertation will present a set of designs including indoor autonomous obstacle avoidance,target tracking and autonomous positioning system based on multi-rotor UAV and the corresponding experimental flight test will be also carried out.The main contents are listed as:Firstly,according to the functional requirements,the hardware design scheme of multi-rotor UAV indoor autonomous obstacle avoidance,target tracking and autonomous positioning system will be presented.Then the hardware and scheme of flight control module,obstacle avoidance module,visual tracking module and autonomous positioning module will be also determined.Furthermore,the obstacle avoidance method of UAV based on potential flow will be put forward to make the UAV smoothly circumvent obstacles in the process of tracking or traveling to the destination.Secondly,an airborne target tracking module based on HOG+SVM detector and KCF tracker will be designed.After the algorithm will also be used to detect,recognize,extract and track the moving objects,the relative position of the moving target is provided for the multi-rotor UAV in real time,which can guarantee that the multi-rotor UAV tracks the target stably.Thirdly,an autonomous positioning module based on in-depth learning and SLAM technology will be designed to provide the navigation and positioning function for the indoor UAV.A deep learning detection point algorithm based on SuperPoint will be proposed to replace the front-end feature point detection algorithm in VINS-Mono to optimize SLAM navigation and positioning.Finally,the feasibility of the autonomous positioning system can be verified.Finally,the experimental results will show that the system of multi-rotor UAV autonomous obstacle avoidance,target tracking and autonomous positioning has good real-time performance,rapidity and stability,which guarantee the important theoretic meanings and application prospects.