Semantic assessment of similarity between raster elevation datasets

Abstract

This paper describes a method to assess the similarity between digital elevation models (DEM), based on the comparison of the landforms. The method attempts to mimic the one commonly used by human beings, which consists of comparisons among the shapes that a human subject identifies in the landscape. To do so, semantic similarity measurements are applied over a hierarchy of concepts. Our method is composed of two stages: the Geomorphometric Analysis and the Semantic Analysis. The first stage aims to represent the topographic properties using one of the concepts of the hierarchy, depending on an analysis of the DEM. The second stage consists of comparisons among the concepts that characterize the landscape using a measure of semantic similarity. In this stage, two levels of semantic analysis are defined: local and global. The advantage of our method is that the interpretation of the results is simplified by means of a semantic processing.