Visual odometry using optic flow for Unmanned Aerial Vehicles

Use of computer vision on Unmanned Aerial Vehicles (UAV) has been a promising area of research given its potential applications in exploration, surveillance and security. Localization in indoor, unknown environments can become increasingly difficult due to irregularities or complete absence of GPS. Advent of small, light and high performance cameras and computing hardware has enabled design of autonomous systems. In this paper, the optic flow principle is employed for estimating two dimensional motion of the UAV using a downward facing monocular camera. Combining it with an ultrasonic sensor, UAV´s three dimensional position is estimated. Position estimation and trajectory tracking have been performed and verified in a laboratory setup. All computations are carried out onboard the UAV using a miniature single board computer.