Measurement uncertainty model for relative visual localization of UAV by a monocular camera

This paper proposes a preliminary uncertainty model for the relative localization of Unmanned Aerial Vehicles (UAVs) using Visual Odometry (VO) based on a monocular camera. The proposed model is based on the 'Guide to the expression of uncertainty in measurement'. In particular, the uncertainty sources affecting a VO-based navigation system are identified, and then a preliminary model considering only the disparity map as an uncertainty source is proposed. Three feature detection algorithms are compared in terms of uncertainty according to the proposed model and by means of a UAV flight simulator. The obtained results demonstrate that the feature detection algorithm proposed by Harris, [12], outperforms the others in terms of uncertainty of the displacement measurements and position estimates.