1gulation.
pp. 225 -
1gulation.
ted Aerial
Annual
nentation
ngulation
p. 644 -
utomated
Digital
ISPRS
g der
1 durch
Heft 392,
3D-CITY MODELING WITH A DIGITAL ONE-EYE STEREO SYSTEM
Felicitas Lang, Wolfgang Forstner
Institute für Photogrammetrie, Universitat Bonn
NuBallee 15, D-53115 Bonn
Tel.: (228) 732901, Fax: (228) 732712
e-mail: Felicitas.Lang/Wolfgang.Foerstner@ipb.uni-bonn.de
Commission IV, Working Group 4
KEY WORDS: Building extraction, City-Models, 3D-GIS, CAD, Digital Photogrammetric Systems, Semiautomation
ABSTRACT
3D-city information is crucial for a number of applications in city planning, environmental control or for telecommunication.
We describe a semiautomatic system for acquiring the 3D-shape of buildings as topographic objects. Buildings are either
modeled as a freely structured union of basic shape primitives or as prisms with an arbitrary ground plan, covering a large
percentage of existing buildings. Interaction takes place in only one image, requiring the operator to specify the approximate
structure and shape of the buildings. 3D-reconstruction including both, height determination and form adaptation, is performed
automatically using various matching tools. The paper describes the features of the system and reports on its efficiency based
on an extensive test.
1 MOTIVATION
There is an increasing demand for three-dimensional (3D)
building extraction as such data is needed for a large number
of tasks related to measurement, planning, construction, en-
vironment, transportation, energy, property management and
transmitter placement. However, systems being capable to
capture entire cities in a relatively short time, at a reasonable
cost and with topographic accuracy are not available. The
reasons are manifold. One certainly is the lack of efficiency,
or the costs for data acquisition, which are higher by a fac-
tor 2-5 than comparable conventional 2D-data. But also the
diversity of users and the resulting lack of clear specification
for 3D-data hinder the developments of operational systems.
Even if such systems would exist, practitioners would hesitate
to acquire such data without having a clear cut application,
as current geoinformation systems do not support the man-
agement of 3D-data.
Though analytical film based systems to a certain degree can
be used for 3D-data acquisition of buildings, efficiency can
only be increased by introducing automation, i. e. image
processing tools into the systems. While orientation algo-
rithms in Digital Photogrammetric Systems (DPS) are highly
developed and the production of digital orthophotos and
the (semi)automatic acquisition of Digital Elevation Models
(DEMs) [Krzystek, 1991] becomes a standard in Photogram-
metry, nearly no support for the extraction of cartographic
features is provided by DPSs. Intensive support of the oper-
ator by automatic tools for mapping will highly increase the
acceptance of DPSs due to the overall higher productivity.
For complex city structures fully automatic approaches for
cartographic feature extraction are of high interest. The state
of the art of automatic and semiautomatic approaches has
been presented at the Workshop in Ascona [Grün et al., 1995]
(see also [Liedtke et al., 1991], [Burns and Riseman, 1992],
[Price and Huertas, 1992], [Bignone et al., 1996]).
These procedures, however, do not seem to be operational
yet. The main reason is the lack of models for linking the ob-
ject classes (buildings, roads, vegetation or rivers) with their
appearance in aerial images and the lack of strategies for
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
bridging the gap between the iconic low level description, i. e.
the image raster, with the semantic high level description, i. e.
the concepts of the cartographic objects in concern. There-
fore, the present goal of most developments is to achieve
partial automation by matching and mensuration tools and
leaving the application dependent decisions to the human
operator (cf. [Quam and Strat, 1991], [Mundy et al., 1992],
[Leberl et al., 1994], [Heuel and Nevatia, 1995]). This al-
lows a smooth transition to more and more automated so-
lutions.
The semi-automatic approach for model-based building
extraction from digital images, was motivated by these
needs. A first version of the system has been described
by [Lang and Schickler, 1993] and [Lang et al., 1995]. It
has been extended by the volumetric modeler (cf.
[Englert and Gülch, 1996]).
The system is based on the following rationales:
e CAD-type modeling for easing link to thematic at-
tributes. ;
e Attributed Boundary Representation for supporting
interaction and visualization.
e Multiple Image Evaluation for increasing complete-
ness , flexibility and reliability of data acquisition.
e One-Eye Stereo for easing interaction and increasing
acceptability by non-experts.
The system is part of our research on automatic
building extraction (cf.[Sester and Forstner, 1989],
[Forstner and Pallaske, 1993], [Braun et al., 1995],
[Brunn et al., 1995]). The system serves two purposes:
e |t is a test bed for the developments in automatic
image analysis, especially for integrating automatic
image analysis procedures.
e It is a practical tool for 3D-data acquisition which al-
lows to analyze the interface between automatic pro-
cedures and operator interaction on large data sets.
In the following the system is described in more detail (section
2). The results of empirical test are given in section 3.
415
———
Nes à
EUM AERA
