Vision-based UAV navigation aided by terrain elevation map

Abstract

A vision-based navigation system with use of terrain elevation map is addressed for autonomous navigation of unmanned aircraft with monocular camera as the supplementary sensor. The aim of this article is to address vision-based navigation techniques for small, low-cost unmanned aircraft, with an emphasis on methods of how to use DTED in vision-based navigation. In this research, new approaches to vision-based navigation with a monocular camera as an alternative to GNSS will be presented. In order to mitigate error accumulation witnessed in vision-based navigation, DTED will be exploited as additional information. Two methods are proposed to combine DTED to vision-based navigation algorithms. First, the vision sensor is used to produce not only the translation by constructing homography relationship, but also height measurement, from aircraft to ground, for referring DTED as done in TRN algorithms. Second, a typical EKF model of bearing-only SLAM technique is examined and modified to add range measurement by using LOS (Line Of Sight) intersection. Simulation studies are conducted to verify proposed methods. The vision sensor is expected to act as an attractive alternative or supplement to navigation systems when the GNSS is denied.