Keeping a moving target within the field of view of a Drone's onboard camera via stochastic estimation

The use of Drones in areas such as cinema, sports, social events, and even video selfies has been increasing due to their flexibility to capture video in scenarios where there is an interest to keep a target within the field of view of the drone's onboard camera. In order to remove the dependence of the pilot that controls the drone, in this work we present a system for an autonomous flight control of the drone with the goal of keeping a target within the field of view of its onboard camera. For the latter, the images were captured by a camera onboard the drone, whose output was combined with a stochastic estimator of the target states, based on the Unscented Kalman Filter, to generate control commands in order that the drone which performs such recording autonomously. The system was validated with real-time tests involving different targets moving with different trajectories and it was compared against human pilots. Our approach kept the target within the field of view with a 96.6% of success compared to 83.3% of success obtained by human pilots. The latter indicates that our approach has the potential to be used in applications where autonomous drones could be used for aerial video recording, with a special interest in keeping a target within the field of view of the drone's camera.