Abstract
This paper presents an effective egomotion approach based on high curvature features extracted from moment images and described using local descriptiors from the original images. Using a least square solution, derived from unit quaternion representation, we apply random sample consensus for motion estimation and show acceptable results without the need for filtering. The six degrees of freedom inherent in the mobile platform allows implementation to a range of autonomous navigation systems. The quality of the obtained results is shown to be comparable with accurate global positioning system-corrected inertial navigation system even for long-range trajectories.
Acknowledgment
The authors would like to thank European Space Agency for their financial support and their technical discussions related to this work.