/ienna; 
and 
be 
CAD 
ate); 
ff the 
9ecial 
E 
diate 
data 
ation, 
s the 
| the 
itives 
alistic 
atural 
bined 
t the 
dered 
s an 
high 
nodel 
ts of 
large 
n of 
ng of 
| yet 
sition 
s of 
ower 
ed to 
scale 
t al. 
three 
e The acquisition of texture data needs to be 
improved. We propose to develop methods 
and instruments to create strip images of 
facades and other objects; 
e The extraction of shape information from 
image data and other source data like the 
exisiting GIS has to be implemented and 
adapted also for the automated reconstruction 
of details like chimneys,  balkonies, bay- 
windows etc.; 
e The processing and enhancement of 
phototexture is in need of significant 
improvement. Shadows, occlusions and 
change in hue and intensity occur and need 
an at least semi-automated solution. 
The current status of these demands is very poor 
and we have insufficient tools to create high 
quality city models in an acceptable time and for 
acceptable costs. For sure a broad field of 
research and development challenges our skills. 
Concept of a Digital Workstation 
In line with the demand for automation mentioned 
above we present the concept of a digital 
workstation which shall provide all necessary 
functions needed to create three-dimensional 
models of cities. We include into this workstation- 
concept also the process of data acquisition and 
the maintainance, update and use of the data. 
Especially for remote use and effective 
distribution of the digital data we have to keep a 
powerful computer environment in mind. We 
present the concept of this geo server. 
The main parts of the workstation concept are : 
e Data acquisition of source data like aerial 
photographs, terrestrial images taken from 
street level and additional data sources like the 
existing 2D GIS or hight information via a 
DTM; 
e Data base content creation which shall lead 
into a fully three-dimensional CAD model 
including surface information via photoexture; 
e Data management in order to maintain, update 
and distribute already existing data; 
e Communication Technology, which prepares 
the necessary background for quick and full 
data distribution. 
We discuss below these prominent parts of our 
digital workstation's concept in more detail and we 
provide also some aspects of modern usage of 
,CyberCity* models. A overview of our workstation 
concept may be seen in Table1. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
Part 1. Data Acquisition/Data Base Content 
Creation 
The objective is to design and develop a low-cost 
collection system for terrestrial | images 
complementing today's collection systems based 
on aerial photography and to develop a 
methodology for producing 3D urban data bases 
with a high level of automation by fusing all 
available sources of information (aerial images, 
GIS data, DEM data and terrestrial images). 
The data acquisition approach will be integrated in 
a vehicle-based kinematic imaging system, based 
on electronic CCD sensors. High resolution image 
data of buildings facades will be produced in order 
to allow geometric and radiometric exploitation for 
a facade's 3D geometry and phototexture [cf. 
Maresch et al. 1996]. In addition to the vehicle 
based sensor system we propose to use a 
rotating camera to map backyard or indoor scenes 
[cf. Duracher et al. 1996]. 
The fusion of all available data sets (aerial 
photographs, digital map, DEM and image data 
from the kinematic imaging system) shall provide 
a high level of automation. Basically two 
complementary approaches will be pursued, each 
based on a two step strategy : firstly, estimate a 
CAD model of the building by using aerial images 
and the knowledge of a building's footprint by the 
GIS [cf. Pasko et al. 1996] secondly, the 
terrestrial images will be used to refine the 
geometry and, mainly, to give the texture 
information on the building's facades. 
Beside these data acquisition and geometric 
processing tasks the digital "workstation has to 
offer functions and techniques to improve 
photorealistic simulation. These thematic 
processing tasks have to be based on manual 
texture processing (like Photoshop of ADOBE) but 
shall rise into innovative techniques to determine 
surface properties from the images, process 
shadows and occlusions in a semi-automatic or 
automatic manner and allow a very effective 
compression of the texture data using parametric 
description approaches. This includes concepts 
for the automatic or semiautomatic segmentation 
of facade texture, where the base texture easily 
can be reduced to a parametric description and 
interesting details are kept in full resolution. 
Finally we have to consider the necessary 
correspondence of 3D geometry and 
photorealistic texture. The need of an accurate 
set-up of source images and the relation between 
CAD-model, phototexture and true object needs 
functions which are well known and are integrated 
in the concept of a Digital Photogrammetric 
Workstations (DPWS). 
Part 2. Data Management 
The data of a three-dimensional urban data base 
need to be managed in line with the evolving 
strategies of three-dimensional geoinformation 
systems (GISs). Various criteria have to be 
fullfilled to allow spatial analysis, up-date and 
maintainance of data and to create subsets of the 
data to meet a user's special requirements.