A Unified Method for Vision Aided Navigation of Autonomous Systems

摘要

In this paper we present a unified method for vision-aided Navigation applied to autonomous systems. Traditionally, vision aided navigation methods can be broadly classified under two approaches: (a) methods that relies only on features available in the scene, and (b) methods that uses registered models of the environments. The first approach employs features tracking that provides only partial navigation aiding. In order to achieve full localization, the second approach that uses registered models of the environment is needed. These environment models could be derived from man-made fiducial or structures of natural terrain. The main drawback of the disparate approaches is due to the lack of navigation solution that is scalable under different conditions.