UAV attitude and position estimation for vision-based landing

Unmanned Aerial Vehicle (UAV) landing safely is key and difficult on UAV research. In order to improve the security of UAV landing on the flight strip, a vision-based method was presented for attitude and position estimation of it, based a few feature points distributed on the ground arbitrarily. Calculating the attitude and position relative to runway is the core. Firstly, the coordinates in the camera frame of the feature points was acquired with the N-point algorithm; secondly, with the orthogonalization method, the rotation matrix and the translation vector between the camera frame and the runway frame were acquired. For the imaging process of feature points is influenced by noise, the least-median-squares (LMS) algorithm was employed to diminish the influence of noise and to improve the robustness. For improving the efficiency of LMS, the data segmentation technology was used. The simulation indicated that the presented algorithms meet the precision demand of UAV landing.