Vision-based Ground Target Tracking Using A Qball-X4 Quad-rotor UAV

Quad-rotor UAV has the unique ability of VTOL (vertical take-off and landing) as well as hovering and cruising at low speed. These abilities give Quad-rotor UAV an obvious advantage in carrying out the missions of searching, tracking, reconnaissance, rescue, and etc. With the advancement of airborne vision technology, the vision-based tracking approaches for moving target by an UVA have very great practical value in various areas. This thesis has developed a complete system of flight control based on visual information. The QBall-X4 quad-rotor UAV developed by Quanser Inc. of Canada is the main hardware experimental platform and QuaRC is the main software environment. The system is developed by Matlab/Simulink, which excels in transformation between real-time control and simulation. It is more convenient for testing and improving control methods. The main duties and contributions of this thesis are summarized as follows:(1) Build a mathematical model of QBall-X4 quad-rotor UAV and build the Simulation model which can be used to simulate at Matlab/Simulink. The double closed loop PID controller is designed according to the structure of the model. Optimize the parameters of PID controller for ITAE target and get a group of optimal parameters. In consideration of the errors that may occur in measurement process and the lost that may happen in the network communication, a Kalman filter is designed to estimate the target motion trajectory.(2) In consideration of the UVA load capacity and image quality, a visual platform on the QBall-X4 quad-rotor UAV is set up. Implement the research on the detection and tracking strategy based on vision information, including Lucus-Kanade optical flow algorithm and random forest algorithm. Due to the mistakes occurring in the process of tracking and detection, this thesis introduces the P-N learning strategy to supervise this process and update the target model to reduce the error probability.(3) Design the flight control system of QBall-X4 quad-rotor UAV for target tracking in the QuaRC software environment. The system includes positioning module, communication module, trajectory tracking control module and hardware drive module and does the experiment of tracking flight in Quanser semi-physical simulation platform. The results of the experiment verify that the system can realize the UVA tracking the target in real-time.