Vision-based UAV pose estimation

Abstract

The popularity of Unmanned Aerial Vehicles(UAVs) has increased dramatically in the past few decades. The main challenging task for UAVs operation is to find their position and orientation accurately and continuously during the flight, using the on-board sensors (camera in our case). There are lots of different approaches for UAV pose estimation in literature now. Several different approach and their problems in real-life application were studied in this research. We chose the real time single camera SLAM method, because our robot was equipped with an on-board camera, and MonoSLAM had proved its reliability in previous localization experiments. In our work, we obtained the ground truth by the Vicon cameras. The experimental results suggest that this method can be reliably used in the control of indoor UAV navigation.