Localization Based on the Gradient Information for DEM matching

This paper proposes the localization algorithm that estimates a ground position by comparing the recovered elevations estimated from aerial images with digital elevation model (DEM). The proposed algorithm consists of two stages: recovering the sampled elevations from multiple aerial images and matching them with DEM. While conventional algorithms estimate the elevation field, that is, recovered elevation map (REM) over a whole image, the proposed algorithm recovers the elevations only at finite number of sample points from a multiple image sequence and does not require rotation of REM. So, the proposed algorithm can estimate the ground position accurately by using a wide recovered area and can estimate the position much faster than conventional ones. Additionally, the proposed algorithm makes use of the gradient information of terrain at multiple sample points of multiple aerial images for considering global characteristics. Computer simulations with various images show the effectiveness of the proposed algorithm.