d 
le 
8 
e 
d 
DO u--— UC. 
   
   
     
face can also be part of a surface object, surfaces of solid 
objects such as building facades and roofs can be attributed 
textures. These can be synthetic ones from a material library 
for photo realistic visualization (near photographic quality) 
and/or rectified photographs for photo true (photographic 
quality) visualization. Decomposition of solid objects into 
boundary primitives serves both the analysis of topology and 
ray tracing. 
Belongs to | Belongs to | Belongs to |Belongs to 
Surface Cas, 7^ TON 
Object ject J Has 7 \_ Object 
Nutt a Nor $m 1 f b us Rt 
(Texture) 
Partof <—7 | Part of | 
     
  
| Represents 
  
Is in 
Figure 1 FDS offering 3D topology. 
The 3D-FDS is upward compatible with a more universal 
data model, the simplicial network data model (see Pilouk 
1996a). This model can also handle objects with 
indiscemible boundaries and facilitates spatial interpolation. 
Such a model is needed for integrated analysis and 
simulations as required, eg, for environmental studies. 
3. PHOTOGRAMMETRY 
3.1 Data Sources 
Data acquisition is one of the major bottlenecks in building 
and maintaining an urban 3D-GIS. Cost effective systems are 
only likely if existing data are incorporated, different data 
sources (existing 2D municipal information systems, 
cadastre, aerial and terrestrial photogrammetry, GPS and 
other ground surveys) are used, automatic 3D model 
construction from disparate data sets is accomplished, and 
if progress is made in automating information extraction 
from images. 
Photogrammetry tenders attractive means for 3D data 
acquisition, specifically in urban areas. Aerial 
photogrammetry 
® is economic for surveying large areas, 
© is flexible with respect to the level of detail and accuraçy 
with the potential of very high accuracy of object 
reconstruction, 
€ can attain a high level of completeness, 
® does not pose problems related to safety in measuring 
urban objects, 
€ allows for a timely survey, 
861 
® grants convenient verification and interactive database 
updating by superimposing vector data on the 
stereomodel, 
€ offers the additional benefit of an image archive and the 
possibility of photo texture mapping. 
Nevertheless, occlusions cannot completely be prevented 
in densely built up areas, even with a good choice of survey 
camera and appropriate flight planning. Moreover, standard 
aerial photographs will not disclose sufficient detail for fancy 
image mapping of facades. Hence (field) completion is 
required, be it through airborne oblique and small format 
photographs, terrestrial photographs on photo-CD (Hóhle, 
1995), digital images from a 3 line CCD camera (Maresch, 
1996), panoramic images from a fish-eye lens (FRANK, 1993), 
video, metric terrestrial photographs or ground surveying for 
detailed reconstruction of facades. The possibility of texture 
mapping instead of surveying facade detail offers cost 
effective information production, comparable to producing 
photomaps instead of line maps in traditional topographic 
mapping. For many applications a visual impression of 
facade and roof detail suffices. This does not hold, eg; for 
3D models for routing of overhead utility lines where the 
relief of facades and roofs must be represented in 3D 
geometry, sometimes with high accuracy (see Baucic, 1995). 
Focusing now on aerial photogrammetry, it can produce 3D 
vector data of topographic objects, ground elevation data 
(thus DTM), and photo textures--especially for roofs and 
terrain, and when colour photographs are available. 
3.2 Tools and Procedures 
3.2.1 Geometry 
Measuring by photogrammetry always has been done in 3D, 
but for a 3D description of topographic objects instead of a 
2D map as a result, we explicitly need all faces, eg of a 
house, its roof facets, its walls, its footprint, and these 
elements in terms of coordinates and topology. This implies 
considerably more work in measuring and structuring data 
than traditional map compilation. Manual data acquisition, 
therefore, must be designed so that only a minimum 
number of points are measured and that the subsequent 
model construction (je object reconstruction and 
structuring according to FDS, thus creating the 3D database) 
can be done automatically. This aim can largely be achieved 
by defining an adequate coding system and a strict digitizing 
procedure (see Wang, 1994). Extracting features and coding 
them in such a way that 3D objects can be assembled, can 
be accomplished by using standard photogrammetric 
digitizing software (eg; CADMAP, KORK, Microstation). 
Algorithmic assemblage of objects from the coded point 
measurements, structuring the data to FDS and checking the 
consistency requires development of software. If, however, 
model construction does not need to result in an FDS 
database, then it could be programmed (or done manually) 
in a commercial CAD package such as AutoCAD or 
Microstation. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996