The main topic of this master thesis is development of a quadcopter which can estimate its position and orientation, thus allowing autonomous flying and performance of different tasks. Structure of the quadcopter is defined and all the elements, which are usually part of its structure, are described. Selection and features of the elements which were used for building our test quadcopter are described in the context of this thesis. There is an overview of all included sensors. Based on the sensors used, methods how to estimate the position and orientation of the quadcopter are presented. The selection concept of hardware and software integration including the architecture of quadcopter control is presented. The calibration procedure of the camera and the way of defining intrinsic parameters of the camera are presented. A large part of this thesis deals with the development of a computer vision application which was used for marker recognition and subsequently for pose estimation based on this information. In this context, all the algorithms which are usually used in computer vision applications are described. Sensor data and pose estimation from computer vision are eventually fused by means of the Kalman filtering – as a part of this, probability theory and algorithm of used filters are presented. There is a presentation of the results of localisation using the Kalman filter and a description of the procedures used for improving localisation. At the end of this thesis the method, how stabilisation and control of the quadcopter based on pose estimation from markers was developed, is presented.
|
