Hand Gestures-Based Smooth 3D Trajectories Computation Applied to Real-Time Drone Control by Tracking 2D Hand Landmarks

Robotic systems are increasingly being adopted for filming and photography purposes. In fact, by exploiting robots, it is possible to program complex motions, achieving high-quality videos and photographs. The contribution of the presented paper is an alternative approach for robot motion planning w.r.t. the joystick. Starting from a state-of-the-art hand tracking system, to detect 2D hand reference points, we built a pipeline to reconstruct 3D trajectories. To evaluate the proposed remote control approach, a captured trajectory is tested in a simulation environment to control the motion of a drone. In addition, experiments are provided using the DJI Ryze Tello drone to prove the feasibility of the approach in real conditions.