Vision-Based Trajectory Tracking System for an Emulated Quadrotor UAV

In this paper, a vision-based system is proposed, using the KINECT sensor (KS), for three-dimensional (3D) trajectory tracking. A target object to track in the 3D space is defined in order to describe its trajectory on a virtual reality environment by an emulated quadrotor unmanned aerial vehicle (UAV). The 3D position vector for the centroid of the target to track is recovered from image processing and information from the depth camera. A color-based object tracking approach is used. A stable fuzzy controller, as well as a LQR controller, for the emulated quadrotor UAV are implemented and compared, where the estimated 3D position of the target is given as reference signal. Important benefits of the proposed object tracking method are its efficiency in the use of the information provided by the KS with a low computational cost for image processing to estimate the target object position and the possibility of improving robustness of the image segmentation by online adjustment of the color filtering parameters. The main contribution of the paper is the use of the estimated position as the reference for the simultaneous control of attitude and translation of a quadrotor from a single controller. Experimental simulation results validate the proposed approach.