Digital Terrain Model Generalization for Multiscale Use

Abstract

Geomatics techniques and applications to process lidar data, large-scale map, stereo-pair airborne photos, and Very High Resolution satellite imagery allow building very detailed Digital Terrain Models. Indeed, in studies characterized by a smaller reference scale, lower resolution models are required to handle a smaller amount of data. Therefore, rather than producing more models with different resolutions, it is preferable to create only one of the multiscale types by using generalization techniques. Different approaches are described in literature in order to achieve this purpose and the results are different in relation to the technique used. This paper aims to compare different algorithms and procedures for Digital Terrain Model generalization. The area selected for this study presents a variegate zone with variable slopes, in order to examine the generalization process in different gradient ranges. Elevation data are extracted from 1:5,000 scale mapping and processed with Geostatistical Analyst to produce Digital Terrain Models with 4 m cell resolution. Five different approaches for generalization are adopted and compared: two based on filtering algorithms (respectively media and median), three on regeneration of Digital Terrain Model interpolating contours or elevation points extracted from the starting model. A new index is provided to evaluate each resulting model also in reference to its capacity to preserve the initial significant values. All the operations are carried out using the Geographic Information System software.
Copyright © 2020 Praise Worthy Prize - All rights reserved.