Complete pose determination for low altitude unmanned aerial vehicle using stereo vision

Abstract

A well-developed pose estimation scenario suitable for low altitude unmanned aerial vehicle (UAV) is proposed. By employing dual CCD cameras onboard, the instant pose of UAV can be determined without any use of expensive sensor like gyro. The unscented Kalman filter (UKF) is hereafter introduced to resolve the highly nonlinear system dynamics as well as the measurement process of the pose estimation system. The only measurements recorded are those snapshots of ground targets/landmarks taken by two CCD cameras. The proposed scenario can also detect large angle rotation of UAV. Simulation is conducted via a simple case, both UAV and ground targets are stationary, to show the feasibility and applicability of the proposed scheme. Actual GPS measurement data of ground targets coordinates was recorded for UKF processing. A highlight phenomenon, implied by simulation, reveals that a sudden transition of estimation errors arises at the epoch when the UAV is experiencing a large angle maneuvering up to 180/spl deg/ per second.