Vision-based state estimation of an unmanned aerial vehicle

Background and Objective: Unmanned Aerial Vehicles (UAVs) have found widespread use in many applications due to its mobility and maneuverability. An important aspect in controlling the movement of these vehicles is its state estimation. State estimation is especially challenging for indoor applications, where Global Positioning System (GPS) signals are weak and have low accuracy. Methodology: This research proposed a vision based state estimation that is applicable even for indoor use. It is a low cost, low power and reliable state estimation approach using a monocular camera with a series of fiducial markers. When a marker is captured by the camera, its position and orientation with respect to the camera's coordinate frame is determined based on its homography transformation. The pose of the camera and hence the vehicle, in world coordinate can then be inferred from known markers poses. Results: In this study experimental results showed that the proposed method is suitable for indoor navigation of unmanned aerial vehicles. The reliability of the state estimation was improved by increasing the number of markers captured. Conclusion: The experimental results verified that the vision based state estimation method for indoor UAV navigation a promising solution and had several advantages over traditional other methods. © 2017 Lee Kian Seng etai.