Camera Path Reconstruction

Abstract

 

Monocular visual navigation is a key problem in the emerging field of Micro-UAVs (unmanned aerial vehicles).  We describe a novel solution to this problem in the calibrated setting, and show results demonstrating its potential.

As required by the application, our solution is unconstrained in terms of camera motion – six DOF (degrees of freedom) are supported.  Moreover, the solution makes no assumptions regarding the content of the scene.

Our solution is of low-level nature, and falls into the category of Visual Odometry.  As such, it may be supplemented naturally by a high-level navigational algorithm involving scene understanding.