| Recently, multirotor UAVs are widely applied in movie, transportation and agriculture industries for its maneuverabiltiy and stability. In order to realize fully autonomous fly, the multirotor UAV needs not only to fly along specified waypoints,but also to land precisely. GPS is effective when the multirotor UAV is in the air,but often fails to fix the position when it is landing near buildings because signals are blocked.A vision positioning approach is proposed to solve the problem in this thesis.This approach adopts a camera onboard to search the center of a special guiding pattern and a sonar sensor to measure height of the UAV. Then position of the UAV can be calculated according to the projection principal.Firstly in this thesis, position measurement of the UAV is studied. A pattern with serval coincide black rings and corresponding feature extracting algorithm are discovered. The equation for calculating relative position is derived from the projection principal. In order to take the inclination of the camera into account, an attitude estimation algorithm is introduced. Then noise filtering and control is studied. Dynamics of horizontal and vertical motion of the UAV is analysed and state space models are obtained. Kalman filter is applied to get rid of noise in the measurement data. Position of the UAV is stablized by PID control law and height controller is designed by a feedback linearization method. Lastly, a model of this system is made. The hardware of this system is consist of a 32-bit microcontroller and an image sensor with array size of 640x480. Algorithms of object searching,Kalman filter and PID controller are implemented in C language. Several programs runing on desktop PC for monitor and adjust are developed.Experiment results show that the UAV can be stabilized laterally and vertically and it’s able to land precisely on the reference point which meets the goal of this thesis. |
PDF Full Text Request