A multiscale approach for terrain referenced underwater navigation

Abstract

A terrain-based underwater navigation using sonar bathymetric profiles is presented. It deals with matching high resolution local depth maps against a large, on-board, low resolution reference map. The matching algorithm locates the local depth map within the a priori larger map to determine absolute position and heading of the vehicle. It uses cliff maps which are steep gradient contours extracted from both local and reference maps. This segmentation provides us with a means to extract critical points, which are defined as high curvature values. The problem is reduced to a singular point-based matching algorithm: given two point sets, find correspondences and estimate transformation between the two sets. In order to register maps at different scales, a part of this study focuses on partial differential equations based scale-space techniques. This approach is tested using real terrain data of the Var underwater canyon (France).