The method presented here, uses three major 
groups of data: 
e Digital Terrain Data (i.e. DTM) 
e Geometric 3D objects, e.g. buildings 
e Man-made objects on the surface, e.g. 
passways, patches, etc. 
2.1 DTM 
Two basic methods can be used for the 
representation of the relief, namely a regular grid 
(Raster DTM) and Triangular Irregular Network (TIN). 
Generally speaking, the grid presentation describes 
the terrain at a regular grid with even distances 
between mashes. This will cause smoothing effects 
and may be insufficient at large scale applications. 
However, the raster-grid presentation allows an easy 
way of handling and storing the data. The uniform 
grid has a benefit of offering simple and easy to 
automate methods for creating Levels Of Detail 
(LODs) which is important for fast visualization of 
huge amount of data (cf. [6]). 
Many investigations have been done on 
polygonization of the surface using TIN (cf. [3]). TIN 
gives the best presentation of the relief in case of a 
rough terrain, with many important features. 
Unfortunately, TIN presentation is quite difficult 
structure for updating. Each modification of the 
surface usually requires re-triangulation of the whole 
TIN. Indeed, there are approaches (cf. [8, 11] 
which allow a re-triangulation only in those restricted 
areas where the changes are made. However, in this 
case information about constraining polygons 
should be introduced and maintained. Another 
drawback of TIN is a quite complex manner of 
creating LOD (cf. [13]) and for this reason there are 
still no algorithms and software developed which use 
TIN presentation for a real-time visualization. 
The type of DTM presentation is an essential aspect 
of the 3D construction of city models. Several 
important considerations influence the choice. The 
urban areas, consisting only of terrain features, are 
relatively small compared to the areas outside the 
city. The surface is basically covered with building 
and man-made objects and only small parts of the 
surface are purely natural terrain areas. The 
expectations are that these drawbacks of TIN will not 
disturb the modelling process. The reason is that 
very small areas have to be re-triangulated in case 
of changes of the surface or for applying LOD in 
real time. 
Another significant consideration is the importance 
of the terrain features in the urban areas. It appears 
that parts which are not man-made or man- 
reconstructed objects are of historical or natural 
importance for the town, requires a precise 
description of the surface. 
The reasoning discussed above leads to the 
conclusion that the TIN presentation is more 
312 
suitable structure for the urban areas than the grid 
presentation. Therefore TIN is used as a surface 
description in the approach described below. The 
manner of gathering data for constructing TIN does 
not matter for the suggested method. The data can 
be obtained either from surveying, or 
photogrammetry, or existing 2D maps with 
distinguished elevation data. 
2.2 Geometric objects 
The most important geometric objects in the towns 
are buildings. The major efforts of the researchers 
working on this topic, are devoted to their 
construction from the row data and their storage. 
Two possible ways of describing are: 
e simplexes 
e parametric description. 
Simplexes can be points, lines, faces, tetrahedrons, 
etc. In the common case buildings are described 
using faces. Faces are required for rendering and 
texturing purposes. This method allows topology to 
be introduced and maintained (cf. [11]). The gain of 
using topology is the ability to share metric 
information by using explicitly described spatial 
relationships. Topology ensures geometric 
consistency of the data model. 
The second method assumes some simplification of 
the row data in order to be obtained regular “boxes”. 
Then these boxes can be described using height, 
width and length. Thus for each particular building it 
is necessary to be known one point with its 
coordinates and sufficient parameters for the 
description of the size and shape of the buildings. 
This method reduces to certain degree the amount 
of data which have to be stored in the data base in 
case of relatively simple constructions. However, 
additional computations for constructing the faces 
are required in the process of visualization. 
The information necessary for describing geometric 
objects can be collected from 2D maps (footprints 
of buildings), aerial images (roof outlines, 
chimneys), terrestrial images (facades’ details), 
measurements from surveying, etc. 
2.3 Surface objects. 
Surface objects are all man-made and natural 
features which lie on the surface, e.g. streets, 
paths, passways, etc. The case, when man-made 
objects are included into the DTM, is clear and 
further processing is not necessary. DTMs without 
information about surface features have to be 
merged with additional data from 2D maps or 
measurements. Thus one significant question 
appears, i.e. how to merge the two data sets. The 
slope of man-made objects differs from the slope of 
the terrain and this fact requires a terrain modelling 
to be applied during data merging. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996 
  
3. 
foll 
COI 
COI 
USt 
of 
Th 
get 
ste