For testing the applicability of concepts and available tools 
we have built a prototype 3D-GIS. Chapter 4 provides an 
outline of its components and presents results from pilot 
studies which have been conducted. At this stage the 
emphasis of our studies is on geometry, leaving aside 
semantic modelling which is very much dependent on the 
specific application context of the GIS. 
Based on the problems encountered in the pilot studies, the 
paper concludes with design considerations for future digital 
plotters which should offer higher 3D productivity. 
2. 3D-GIS 
2.1 Requirements 
A 3D municipal information system should support civic 
activities ranging from administration to planning and 
construction for city (re)development, planning and 
management of utilities, security assurance, environmental 
management, and conservation. The promise of 3D-GIS is 
better accessible inventories, extended spatial analysis, 
refined simulations, and more appealing visualization than 
either 'flat' GISs or solitary CAD models can grant. Moreover, 
a 3D-GIS could also support automation of 
photogrammetric object reconstruction by providing 
information for model based image analysis. 3D analysis 
capabilities could also be exploited to efficiently plan and 
conduct field completion (see Felus et al, 1996). 
This broad spectrum of application presents multifarious 
requirements for the urban 3D-GIS these include: 
€ provision of means to construct the digital model 
representing geometric and semantic properties of 
objects from various inputs and handling implied 
uncertainty 
integrity and maintainability of the database 
support of spatial analysis in different user contexts, 
including basic operations such as: 
measure spatial extensions of objects (volume, 
surface area, line length) 
answer topologic queries (pipe passing through the 
building?) 
search and retrieve objects of certain properties 
zoning, route optimization, lighting studies, etc 
support of simulations (eg environmental impact 
analyses) and design (virtual updating of the existing 
situation by designed objects) 
effective visualization (perspective and orthogonal views, 
user (interactively) defined viewing direction, viewing 
point and stretch, photo realistic computer graphics and 
photo true pictures) 
promotion of different abstraction 
resolution) 
levels (multi- 
A discussion of application areas of an urban GIS and 
requirements for its 3D model can be found, eg in (Fórstner 
et Pallaske, 1993). Pilouk (1996b) gives an overview of 
possible system architectures. 
860 
2.2 Data Model 
The analysis of current GIS and CAD capabilities and 
software tools offered (see Bric et al, 1994) let us to decide 
to first look for rigorous spatial concepts for data modelling 
and on such a solid basis, see which available tools can be 
utilized and how. Focusing on urban scenes and 
considering the above requirements, we have selected a 
vector data model based on Molenaar's FDS. 
The comprehensive 3D data model comprises geometric 
and semantic description of objects and allows for the 
analysis of 3D topology. It can be handled by a single 
DBMS. It is downward compatible with 2D-FDS so that data 
from existing GISs can be incorporated after being 
converted to this structure. It follows a 3D boundary 
representation because of the desired visualization 
capabilities. Photogrammetry offers accurate and economic 
techniques for the acquisition of vector data on urban 
topographic objects and provides for the input for mapping 
photo textures onto surfaces of objects. These raster data 
can be integrated into the vector model. Since we are 
mainly interested in object reconstruction, we can limit our 
review of FDS to the representation of geometry without 
elaborating on the modelling of semantics. 
The 3D-FDS groups objects to classes in a thematic sense 
and distinguishes four types of objects according to 
geometry: point, line, surface, and body object (see figure 1). 
The representation of geometry aims at separating its three 
aspects: topology, shape and size, and position. Positional 
information is contained in the geometric primitive node 
with its x,y,z coordinates. The one-dimensional geometric 
primitive is arc. Arcs constitute line objects and boundaries 
of faces. Faces are two-dimensional primitives, they 
constitute surface objects and boundaries of bodies. Line 
objects, surface objects, and solid objects (bodies) have 
shape and size. To avoid ambiguity in the data model, arcs 
that constitute the boundary of a face, must be relatively 
oriented; the auxiliary geometric primitive edge is 
introduced for this purpose. Hence, /eftand right body to a 
face can be defined. The topologic relationships between 
the objects are defined through the geometric primitives 
and the links between them which are shown in figure 1. 
This diagram, together with a set of conventions, represents 
the basic data model. The conventions are rules that must 
be observed when structuring data; eg, faces must not 
intersect or touch, an arc cannot intersect a face, line 
objects have no branches, etc. When introducing the 
conventions that arcs are straight lines and faces are planes, 
they would not need any shape parameters. The data model 
as shown in figure 1 supports the handling of a large variety 
of queries and can be extended to build composite objects 
(see Bric et al, 1994). Moreover, it can be upgraded to 
structurally integrate a digital terrain relief model (DTM), see 
(Pilouk et Tempfli, 1994). 
The 3D-FDS can be seen as an extension of the edge based 
boundary representation used in solid modelling. Since any 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996