Visual odometry and scene matching integrated navigation system in UAV

Based on the real-time image sequence of airborne, a novel vision-based integrated navigation method of UAV which combined the relative and absolute navigation algorithms is proposed. Visual odometry is used as the relative algorithm in order to estimate the position of UAV by the derived homography matrix between two consecutive images. RANSAC algorithm is working to remove mismatching points. To eliminate the serious error accumulation of visual odometry and to improve the positioning accuracy and reliability, absolute method based on scene matching using FREAK descriptor is introduced and to build an integrated navigation system. By the data collected by semi-physical simulation software, the method is tested. The data indicates that the proposed method can achieve relatively high position accuracy and can act as an effective navigation method in UAV navigation system.