ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
GENERALIZATION FOR 3D GIS 
Fengwen BAI, Xiaoyong CHEN* 
Space Technology Applications and Research Program 
School of Advanced Technologies, Asian Institute of Technology 
p.O. Box 4, Klongluang, Pathumthani 12120, Thailand 
E_mail: fwbai64@yahoo.com, xychen@ait.ac.th 
KEYWORDS: Generalization, 3D, GIS 
ABSTRACT 
The development of Geographic Information System (GIS) has inputted new vitality for the map that has been used thousands of years 
as the widespread and common tool for academic and reference object. 3D GIS as the characteristic of visualization of the reality is 
more and more becoming an indispensable component of future's GIS thought automated generalization in 2D digital map/GIS is still 
the outstanding question. The future availability of automatic generalization tools has allowed us to realize our goal, namely the 
generalization in 3D GIS, will be the basic research in the field. The paper which is focused on the theoretical research for 3D GIS 
based on the characteristics of 3D GIS explores several different ways for generalization associated with simple examples: a) Levels of 
systems which displays different detail elements comparing to the area or scale from large scale 3D GIS to small one; b) Symbolization 
in both small and large scale 3D GIS, including the design of symbol; c) Refinement such as selection, aggregation and simplification 
etc. of features are used to generalize features according to the functions and area characteristics. Those will provide an explorative 
approach to the 3D GIS production in normalization and comparability to each other. Some basic algorithms and operators have been 
expanded from 2D GIS to 3D GIS. 
INTRODUCTION 
Traditional maps displaying the Earth in 2D have been used 
thousands of years. Human beings had been not realized their 
dream of displaying the real world as it is in 3D and more until 
the invention of computer. The mapping industry has taken the 
revolutionary move of replacing 2D, traditional manual work by 
3D and computer-driven processes. Generalization, the art 
and theory with which cartographer describes the world by 
selecting and drawing elements on the map, has been 
developed at high level in 2D mapping (It does not overstate 
that a cartographer is an artist). Semi-automated and 
automated generalization theories have been developed and 
used in variable of mapmaking. 
Geographic Information System (GIS) that serves long-term, 
multiple-purpose applications has become a hot industry in the 
recent decades. GIS is as a tool that captures, stores, 
manages, manipulates, analyzes, models and displays 
information with respect to geographic space. It is the rapidly 
growing information technology area at today’s information 
age. 
Generalization is the simplification of observable spatial 
variation to allow its representation on the map (Goodchild, 
1991). It is an information-oriented process intended to 
universalize the content of a spatial database for what is of 
interest (Muller, 1991). Relating observation of geographic 
space to models of geographic phenomena converts data to 
knowledge, Once a model of the phenomenon has been 
developed, it is possible to generalize that model. Cartographic 
generalization may be considered to be an exercise in applied 
geography (Pannekoek 1962); for example, generalization of 
contours, coastlines, streams, and topographic surfaces 
should incorporate knowledge of landforms and geomorphic 
process. Furthermore, the cartographic character of many 
features is scale-dependent (Buttenfield 1989), and the use of 
a single database to support mapping at a variety of scale 
presents many research challenged (Beard 1987). 
Generalization in 3D GIS will become an indispensable 
component of future's GIS. Whatever the task, compiling data 
in different sources into a master database to compose an 
output from it for data compression, analysis, or 
representation, generalization is most likely involved. 
Map generalization. One of the greatest challenges to 
cartographers working with digital systems has been 
automating the map generalization process (Robert, W 1999). 
Manual generalization, as the traditional means, is extremely 
subjective and time consuming. Cartographers draw a reduced 
map by hand. They eliminate unimportant features, simplify 
lines and boundaries, combine area features, and resolve 
conflicts as they draw. The result is, no doubt, operator- 
dependent. As geographical databases are constantly built, 
the request for automation of generalization capability for 
multi-purpose output is in demand. The field of generalization 
has extended to include 3D GIS applications. Noticeable 
efforts have been made by researchers and some 
GIS/mapping software vendors defined digital generalization 
problems and developed solutions. However, none of the 
existing 2D GIS/mapping systems has provided a set of tools 
that fully satisfies digital generalization needs. Aspects as 
automation of generalization are still open problems. 
The age movement forward is carried by both the existing 
problems solved and new matters found. The future availability 
of automatic generalization tools has allowed us to realize our 
goal, namely the generalization for 3D GIS, will be the basic 
research in the field. 
For the main topic of the paper, there are two niches in the 
research and market. 
• In Research 
Up to new, a set of complete theories of generalization in 
traditional map had been created and these theories have 
been transplanted to the digital map and 2D GIS. Automation 
or semi-automation of generalization in digital map and 2D GIS 
has being developed based on these basic theories even there 
are still some open problems needing to be found. 3D GIS as 
the new baby generated with the computer development will 
go along the different way: combination the basic theories with 
automation. Generalization for 3D GIS would be very valuable 
to support the realization of automation. 
• In Market 
Normalization not only for digital map or 2D GIS but also for 
3D GIS is the major question in data product. How to rule a 3D 
GIS product is reasonable or not for its data, detail and 
selection of elements, etc.? A building, for instance, can be 
displayed as a symbol, a sketch with only the walls and the 
root or the almost real one that has windows, doors, even the 
pattern of the wall. Which one is the better to the user and 
which one is fair to the investment. The research can be a 
Supported by Visiting Scholar Foundation of Keb Lab. In Wuhan University, P. R. China