A method for the visual estimation and control of 3-DOF attitude for UAVs

This study describes a novel, vision-based method for automatically obtaining the 3-DOF attitude of an aircraft. A very wide-angle view of the environment is captured and the input images are classied into fuzzy sky and ground regions using the spectral and intensity properties of the pixels. A novel approach to obtaining the 2-DOF attitude of the aircraft from the classied images is described, which is used to generate a stabilised panoramic image of the horizon and sky. This `visual compass' is then used to determine the heading direction of the aircraft. We present results from recent flight testing that demonstrates the abilty of this vision-based approach to outperform an inexpensive inertial unit and provide real-time guidance for a UAV