UAV Attitude Control using the Visual Horizon

We present a technique for measuring, controlling, and stabilizing the attitude of a UAV by using a camera to monitor the visual horizon. A vision-based algorithm incorporating color and intensity information is used to detect the horizon by segmenting the ground from the sky. The attitude of the aircraft is then measured using the position, shape, and orientation of the horizon profile in the camera image. We show that this information can be used to stabilize the roll and pitch of the aircraft, to achieve and maintain any desired attitude, as well as to orchestrate a number of challenging aerobatic maneuvers that are within the limits of the aircraft’s performance envelope.