Position tracking system using single RGB-D Camera for evaluation of multi-rotor UAV control and self-localization

Abstract

In this paper, an evaluation system for verifying algorithm of unmanned aerial vehicle(UAV) control and self-localization in GPS-denied environments, instead of the Vicon motion capture system that is being widely used as ground truth of the quadcopter position control, is developed and its feasibility evaluated. Among the RGB-D cameras that can make 3D measurements, Kinect is used and validated by applying the algorithm to a linear robot, which then chooses a marker and recognizes it. The marker recognition algorithm comprises two parts. First, a background extraction technique is utilized to establish the area of interest; next, a filter that adopts the labeling technique is used to recognize a marker in the region of interest. The feasibility of the algorithm is validated in an actual experiment employing two kinds of quadcopters.