Visual-GPS combined 'follow-me' tracking for selfie drones

The 'follow-me' mode, where the drone autonomously follows and captures the user or target, is a new and attractive feature for camera drones, especially selfie drones. For this purpose, today's commercial drones use the difference between two GPS data in the drone and user-side mobile GCS, e.g. a smartphone, but the targeting performance is often not satisfactory due to the inaccuracy of the GPS data, ranging from a few to tens of meters. Visual tracking can be considered for a solution to this problem, but the reliability of visual tracking is still questionable for long-term tracking in unexpected operating environments. The paper proposes a hybrid approach that combines the high accuracy of a visual tracking algorithm in short-term tracking and the reliability of GPS-based one in long-term tracking. The experiment with our prototype drone system demonstrates that the proposed combined approach can accomplish the follow-me operation very successfully, capturing the target in the center of video contents over 50% higher accuracy than the GPS-based ones. Also, the extreme scenario experiments verify the system can recover vision tracking failure and Wi-Fi failure quickly in a short-term, e.g. 3-7 s.