IN TRANSITION FROM 2.5D-GIS TO 3D-GIS 
Dieter Schmidt and Dieter Fritsch 
Institute of Photogrammetry 
Stuttgart University 
P.O.B. 106037 
70049 Stuttgart / Germany 
dieter.schmidt@ifp.uni-stuttgart.de 
Intercommission WG III/IV 
KEY WORDS: GIS, Three-dimensional, Data Model, Integration, Algorithms 
ABSTRACT: Some of the 2.5D geographic information systems can easily be extended for the third dimension. The ex- 
tension is exercised by supplementing a 3D data model and additional 3D operations. For computing the three-dimensional 
operations one can refer to the alrady existing two-dimensional operations. At the implementation into Smallworld GIS the 
available source code can quickly be extended for the third dimension. The presentation is done by an external 3D graphic 
tool. 
1 INTRODUCTION 
Although our perception is three-dimensional, spatial ob- 
jects in a geographic information system are for the time 
being mostly managed in a two-dimensional way. Therefore 
the present research activities pay special attention to the 
development of a 3D-GIS. There is an existing need for 3D 
data, but the handling is more difficult and requires more 
complex algorithms as it is in the two-dimensional case. An- 
other problem is to get hold of 3D information. 
The three-dimensional topology is already widely used for 
computer-aided drafts or computer animation and, besides 
that, software and hardware offer more and more 3D sup- 
port. The results in the three-dimensional representation of 
real objects stand for the further development of geographic 
information systems with special emphasis on the analysis 
of geometry and attributes. 
A first pragmatic approach of the extension of an exist- 
ing commercial 2.5D-GIS is to attribute faces for additional 
height values. In doing so it is possible to receive a simple 
three-dimensional representation which allows an adequate 
approximation (of buildings for example) for certain appli- 
cations. The representation is then achieved by an external 
3D graphic program. Additional parameters like roof slopes 
and shapes could support the representation and help to 
achieve better accuracy. 
Should one require, however, a more substantial representa- 
tion as for instance in the field of urban planning, then the 
introduction of an additional three-dimensional data model 
cannot be omitted. 
2 THE 3D DATA MODEL 
As already known, two-dimensional data models suffer from 
its limitations to model 3D solid objects. By using digi- 
tal terrain models (DTM) only three-dimensional surfaces 
can be described with single z-values. As to introduce an 
additional three-dimensional data model the definition of 
relations to the two-dimensional geometry and thematic at- 
tributes is necessary. An unique identifier links both the 
geometry model and the attributes. In most cases the data 
model will be stored in a relational database. As demon- 
strated by Molenaar (1992) an analysis can then be per- 
formed through the basic language SQL. 
The definition geo-object used in the following, describes an 
object with spatial relation. It consists of thematic and 
spatial attributes. Spatial attributes can either be two- or 
three-dimensional. In addition to a 3D attribute every 3D 
geo-object holds as well a 2D-attribut which represents its 
outlines, i.e. every 3D object has a two-dimensional mark. 
In case of a small number of 3D objects it does not matter 
developing a 3D access structure, as every geo-object will be 
at least referenced through the two-dimensional index. 
A three-dimensional data model (see fig. 1) was designed in 
the notation of [Rumbaugh, Plaha, Premerlani, Eddy and 
-. Lorensen, 1991] . The rectangles describe object classes (like 
748 
the object class point), lines denote the association between 
object classes. A point at the end of an association line 
marks an 1:n relation. A triangle at an intersection of a line 
marks a generalization (for example node and intermediate 
points are subclasses of the class point). Generalization and 
specialization are the general terms to describe both direc- 
tions of inheritance. A rhomb on the connecting line de- 
scribes an aggregation (for example a point- shaped object 
consists of a node). 
Within the rectangle, which describes an object class, the 
name is written in bold on top, below follows a line. Under- 
neath this line attributes of the class will be numbered. An 
oblique stroke in front of an attribute denotes that the value 
can be calculated and has not to be stored (but will mostly 
be done because of efficiency reasons). If a face is planar 
it will be calculable but a calculation for every examination 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
ex 
CO! 
CO! 
ty] 
tw 
on 
Wa 
wh 
att 
is - 
no 
mo 
to 
WO! 
for 
by 
anc