Circular motion standalone tracking of a target using a quadrotor with one cheap fixed monocular vision camera

• Main Author: Tan, Jialiang
• Other Authors: Wang Jianliang
• Format: Final Year Project
• Language: English
• Published: 2014
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/61483
• Institution: Nanyang Technological University

In this project, a software is developed to enable a low-cost quadrotor with monocular camera vision (Parrot AR.Drone 2.0 in this project) to track and circulate around a moving target object on the ground with steady velocity. The software consists of three major components: a target extraction algorithm for extraction of the target object from the raw image, a target coordinate compensation for compensating the bias that are introduced due to the unstable pose of the quadrotor and finally three PID controllers that produce steering commands. The main portion of this project is the coordinate compensation. The virtual image concept is introduced, and in combination with the inexpensive onboard sensors, the bias was compensated and the quadrotor is able to perform solid visual based tracking in a GPS-denied environment with relatively light and hence fast calculation that does not involve 3D mapping. Experiments have demonstrated that the tracking was robust and the circulation was smooth in terms of the trajectory.