Visual servoing of an autonomous Micro Air Vehicle for ground object tracking

This paper describes an object tracking system using an autonomous micro air vehicle (MAV) and demonstrate its potential use for civilian purposes. The vision-based control system relies on a color and feature based vision algorithm for target detection and tracking, Kalman filters for relative pose estimation, and a nonlinear controller for MAV stabilization and guidance. The vision algorithm relies on information from a single onboard camera. An arbitrary target can be selected in real-time from the ground control station, thereby outperforming template and learning-based approaches. Experimental results obtained from outdoor flight tests, showed that the vision-control system enabled the MAV to track and hover above the target as long as the battery is available.