A DATABASE FOR A 3D GIS FOR URBAN ENVIRONMENTS SUPPORTING PHOTO-REALISTIC 
VISUALIZATION 
Michael Kofler, Herwig Rehatschek, Michael Gruber 
Institute for Computer Graphics 
Graz Technical University 
Austria 
{kofler,herwig,gruber}@icg.tu-graz.ac.at 
Commision Il, Working Group 2 
KEY WORDS: GIS, Urban, Database, Visualization, Three-dimensional 
ABSTRACT 
As 3D GISs (geographical information systems) become increasingly important, new database technologies are needed. This 
paper presents ideas about requirements for a 3D GIS database for urban environments. A high emphasis is given to the 
support of a fast photo-realistic visualization, thus expanding the scope of functions offered by traditional GIS databases. 
Required functions to allow an efficient visualization are: e Perspective querying (to extract all objects of the database located 
a pyramid of vision). e Support of the level of detail concept on database level (to extract objects in varying detail according 
to the size of the objects on screen). e Image compressing/decompressing/caching (to offer a scale-able compromise between 
speed and data volume). 
KURZFASSUNG 
Mit der zunehmenden Bedeutung dreidimensionaler geografischer Informationssysteme (3D GIS) zeichnet sich die Notwendigkeit 
neuer Datenbanktechnologien ab. Herkómmliche GIS-Datenbanken sind insbesondere da überfordert, wo es um die schnelle 
Visualisierung der Daten geht. Erforderliche neue Funktionen sind: e Perspektivische Abfragen (um alle Objekte zu ermitteln, 
die momentan sichtbar sind). e Unterstützung des Level-of-Detail-Konzepts (um Objekte je nach ihrer aktuellen Größe am 
Bildschirm in unterschiedlichen Qualitátsstufen zur Verfügung zu stellen). e Automatische Komprimierung / Dekomprimierung 
von Bildern. 
1 INTRODUCTION / MOTIVATION 2 RELATED WORK 
An increasing number of users of GISs for urban environments This section summarizes some recent papers on rendering 
asks for 3D extensions ([SBM 94], [RG 95]): and storing complex geometrical models. All papers present 
ideas to improve the LOD (level of detail) concept, which 
will be crucial for any 3D GIS application supporting inter- 
active photo-realistic rendering. LOD is a mechanism used 
in computer graphics to improve the drawing speed of com- 
plex scenes [Cla 76]: Each object is stored several times in 
different levels of quality (levels of detail). During visualiza- 
tion each object is drawn in the optimal level of detail. The 
: à chosen level depends on the size of the object in the current 
e Tourism agencies want to offer virtual reality models of View. Objects that appear small can be drawn in little detail 
destinations they are offering. (and therefore very fast) without loosing quality; in contrast, 
The step from 2D to 3D causes several problems, which can objects near the point of view that cover a lot of space on 
be summarized in three points: building 3D models, storing | the screen need to be rendered in full quality. 
e Scientists want to simulate noise, heat and exhaust 
spreading in big cities. 
e Telecommunication companies need 3D data to calculate 
wave propagation in urban environments. 
e Architects want photo-realistic models of existing build- 
ings to plan new ones and to visualize the resulting 
scenery. 
them and providing a user interface to visualize and manip- Although the LOD idea is quite old and lots of applications 
ulate them. This paper is focused on the latter two points. ke use of it, several problems remain unsolved. The most 
It discusses techniques to store and visualize huge 3D data urgent one is probably building LOD models: although sev- 
sets. eral scene editors already support the LOD concept, none is 
Currently, two research groups are working on this topic: on able to build LODs automatically in good quality yet. There- 
the one side, maintainers and developers of GISs are trying fore, most existing LOD models are still built manually, thus 
to expand their 2D systems with 3D features. On the other increasing the already high cost of creating a model at all. 
side, lots of efforts come from computer graphics experts, [HG 94, SS 95] explain algorithms for building lower levels of 
who want to visualize scenes of growing complexity faster detail from a given 3D model in high quality. Until now, these 
and faster. algorithms have been restricted to polygonal models and do 
NS : : i h 
Unfortunately, there is little interconnection between these not deal with textures, though 
two groups. This paper tries to fill this gap. It gives an In its original concept LODs offer different geometric rep- 
overview of recent research activities (section 2), points out resentations for one object. [MS 95] propose two improve- 
requirements to a 3D GIS database for urban environments ments: (1), the LOD concept can be adapted to store the 
(section 3) and discusses issues of implementation (section entire model (e.g. a city) in one hierarchical structure. (2), 
4). The paper ends with a short outlook (section 5) and 'impostors' replace traditional LODs: Impostors are very simi- 
some references. lar to LODs; they are more versatile as they include additional 
198 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996 
info 
pos 
one 
ity 
buil 
[FS 
Thi: 
buil 
cho 
use 
the 
star 
scer 
stor 
dah 
but 
that 
[KL 
fies 
dat: 
tor 
achi 
gual 
is h 
triar 
3.1 
Curt 
data 
spec 
fact 
(00 
The 
to tl 
data 
struc 
such 
obje: 
RDE 
Thei 
of cc 
port 
(e.g. 
othe 
perfc 
In co 
mod 
for ( 
be ci 
binar 
this 
langı 
avail; 
OOD 
a bel 
obvic 
3.2 
In Se 
Was |