Contour extraction from digital elevation models for topographical mapping with minimal information loss

Abstract

This thesis defines a measure of the amount of information lost when extracting contours from a Digital Elevation Model (DEM). The measure is based on triangulation of the contour vertices and Sibson’s natural neighbor interpolant Z(1). The length of the gradients at the triangulation points are derived from the natural neighbors defined by the triangulation itself, clamped by a minimal length. The gradient directions are, by definition, perpendicular to the contours. As contours are used to visualize a surface, it is important that the plot is not overly cluttered and impedes readability. This constraint is added to the optimization problem by defining a measure of the squiggliness of a contour. The measure is inspired by the bending force needed to deform a thin metal rod to the shape of the contour. Different methods are applied to the task of minimizing the information loss and contour squiggliness and their performances are compared. An application is created for DEM generation from Lidar data, contour extraction and visual comparison of the extracted contours. The application aids in parameter tuning for the tunable models. The extracted contours are also visually compared to hand-made contours by professional orienteering mappers.