Comparative performance of two methods for visual terrain assisted navigation

Traditional terrain assisted navigation systems are based on correlating terrain profile measurements obtained using a radar altimeter with a Digital Terrain Map. Recent works have shown that multiple-frames acquired from a monocular camera can also be used in combination with a Digital Terrain Map to compute a navigation solution. This opened the possibility of using terrain assisted methods for small and medium air and underwater autonomous vehicles for which radar or sonar altimeters are too expensive in terms of volume, weight or price. The purpose of this paper is to compare two alternative approaches that have been proposed for solving the visual terrain assisted navigation problem. The first, known as C-DTM, is a two-frame approach that can be seen as an extension to the Epipolar geometry under a planar scene. The second incorporated DTM data as additional constraints into the Bundle Adjustment algorithm, the latter being the most popular way of solving the structure from motion problem in use today.