GENERALIZING RELIEF REPRESENTATION USING DIGITIZED CONTOURS 
Wanning Peng, Morakot Pilouk, Klaus Tempfli 
International Institute for Aerospace Survey and Earth Sciences (ITC) 
P.O. Box 6, 7500 AA Enschede, The Netherlands 
Tel. +31-53-4874358, Fax. +31-53-4874335 
E-mails: PENG@ITC.NL, MORAKOT @ITC.NL, TEMPFLI@ITC.NL 
XVIITISPRS Congress, Commission IV, Working Group 4. 
KEYWORD: GIS, Generalization, DEM/DTM, Algorithms, Vector. 
ABSTRACT 
This paper introduces an approach for the generalization of terrain relief represented by a digital terrain model (DTM). It 
also presents an automatic vector approach to improve TIN DTMs obtained from digitized contours. Important terrain 
skeleton information in the form of point data is extracted automatically to solve the problem of flat triangles. These points 
can also be used to further extract and form the skeleton lines which can be used as constraints in a generalization process. 
Methods to determine the planimetry and elevation of the "skeleton" points are investigated. The algorithm is tested using 
ISNAP, a Windows based software package developed by the authors using the C++ programming language. Examples are 
also given to demonstrate the potential of the proposed algorithm. Finally this paper gives an outlook for further 
development. 
1. INTRODUCTION 
Terrain relief information plays a very important role in 
many GIS applications. Due to the limitations of available 
tools, this three-dimensional information traditionally is 
mainly represented as contour lines in a two-dimensional 
space such as map sheet. As a result, relief generalization is 
normally (implicitly) conducted via the generalization of 
contour lines which is initiated by the need of map scale 
reductions. As a contour line is not a real terrain feature, but 
an (isolated) imaginary line connecting terrain points of 
same elevation, contour maps do not provide immediate 
images of relief characteristics for the readers. Generalizing 
contour lines therefore requires some kind of "imagination" 
that "captures" the relief characteristics of terrain surfaces 
from a set of contour lines that are naturally interrelated in 
a certain way through the nature of terrain relief and 
constraints of man-made features. 
Relief generalization became an apparent subject after 
digital terrain models (DTM) were introduced to represent 
terrain relief since late 1950s. As much of the earth's 
surface has been mapped as contour maps, contour to DTM 
conversion has been a common approach to obtain a DTM. 
A TIN DTM obtained from digitized contours likely 
contains flat triangles. Flat triangles create artificial terraces 
thus provide incorrect information about terrain relief, 
which in turn, will have effects in generalization decision- 
making and contouring. Several approaches have been 
proposed to solve this problem. Manually adding terrain 
649 
skeleton information (e.g. break lines, spot-height) is an 
example that may solve the problem. However, it requires 
special skill and is a laborious approach. "Triangle 
swapping" is another approach, which is limited to the 
places where the two adjacent triangles form a convex 
quadrangle. An efficient method is to automatically extract 
terrain skeleton information from the contour lines based on 
their shapes and patterns. Known existing approaches are 
based on distance transformation, which requires to operate 
in raster domain (Pilouk, 1992, Tang, 1992). The raster- 
vector conversion and vice versa are thus necessary and 
may require manual editing, which implies extra processing 
steps. 
In this paper, we first introduce an approach for the 
generalization of terrain relief representation, based on 
some existing methods, then present an automatic vector 
approach to improve TIN DTMs obtained from digitized 
contours by solving the problem of flat triangles. Critical 
points that represent (or approximate) the skeleton locations 
are extracted as additional information after the first 
(constrained) triangulation has been completed. The process 
makes use of human knowledge as well as information from 
the original contour lines, topology, and the properties of 
geometric elements of the network. Methods to determine 
the planimetry and elevation of the "skeleton" points are 
investigated. The points are then added to the point set and 
inserted into the current model to obtain a new surface 
representation by local updating. These points can also be 
used to further extract and form the skeleton lines which can 
be used as constraints in a generalization process. The 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996