3
2
port
lave
dew
ally
'ood
ects
East
the
et, it
ase.
ig in
nple
and
and
TK-
EM,
king
pace
uced
over
high
tnik,
Al-
ified
„and
for
"inal
AUTOMATED GENERATION AND UPDATING OF DIGITAL CITY MODELS USING
HIGH-RESOLUTION LINE SCANNING SYSTEMS
G. Vozikis
IPF, Vienna University of Technology, Gusshausstrasse 27-29, 1040 Vienna, Austria
gvozikis@ipf.tuwien.ac.at
Commission III, WG 111/7
KEYWORDS: Databases, Remote Sensing, Automation, Change Detection, Extraction, Building, Satellite, High Resolution
ABSTRACT:
During the past few years lots of research has been carried out in the field of building extraction from airborne laser scanner data and
airborne large-scale imagery. This data can be used to create highly detailed Digital Surface Models (DSMs) and eventually Digital
City Models (DCMs). It seems that there exists also a potential in data acquired by high-resolution satellites and airborne multi-line
scanners in order to derive these models.
Already existing extraction-algorithms for Lidar data cannot be adopted to those spaceborne and airborne imageries since the
geometric resolution is too coarse. The proposed method can be divided into the following steps. The derived normalized DSM is
used together with image data to localize potential building areas. Afterwards the building edges and corners are extracted, which is
carried out with an adaptive region growing process and an iterative step-by-step Hough transformation. Finally, change detection
techniques and a method for updating existing data-bases in an automated way are discussed.
KURZFASSUNG:
Wihrend der letzten Jahre wurde viel Forschung im Bereich der Gebiudeextraktion von Laserscanner- und Luftbildern betrieben.
Diese Daten werden verwendet um detaillierte Oberflichenmodelle (DSMs) und eventuell auch Digitale Stadtmodelle zu erzeugen.
Es besteht die Vermutung dass auch bei hochauflösenden Satellitenbildern und flugzeuggestützten Zeilenscanneraufnahmen ein
gewisses Potential zur Erzeugung dieser Modelle existiert.
Bereits bekannte Extraktionsalgorithmen aus dem Bereich der Laserscannertechnologie kónnen leider nicht adaptiert werden, da die
geometrische Auflósung der Satelliten- und flugzeuggestützten Zeilenscanner-Daten zu grob ist. Die vorgeschlagene Methode kann
in folgende Schritte unterteilt werden. Das abgeleitete normalisierte DSM wird gemeinsam mit den Bilddaten zur Lokalisierung
potentieller Gebäuderegionen verwendet. Anschließend werden Gebäudekanten und —ecken extrahiert, was durch ein adaptives
region-growing' und einer iterativen ,Schritt-für-Schritt Hough Transformation geschieht.
Abschließend werden Methoden zur. Veründerungsdetektion und Instandhaltung von Datenbanken vorgestellt.
I. INTRODUCTION
Digital City Models (DCMs) have become one of the most
important and attractive products of photogrammetry and
remote sensing. The spectrum of application areas dealing with
such models is huge: environmental planning and monitoring,
telecommunication, location based services, navigation, virtual
reality, cadastre etc.
During the last decade photogrammetry and remote Sensing
experienced a big technological advance. On the one hand,
because the geometric resolution of satellite imagery has
increased tremendously (footprints of one metre or less are
current state); on the other hand photogrammetry has become
more and more 'digital and new techniques are needed to
accelerate the workflow. Due to this progress the science of
satellite remote sensing starts dealing with application areas,
that until now where classical photogrammetric tasks, such as
large scale mapping or city model generation.
A challenging goal is to automate and accelerate the production
of DCMS, to increase their quality and to keep them up-to-date.
When done manually this procedure is one of the most time
consuming and expensive tasks.
1033
The use of airborne laser scanners and the related processing
techniques were a big step towards faster city model extraction.
Though extensive research has been carried out, (e.g. Brenner,
2000; Haala et al., 1998; Vosselman and Dijkman, 2001) they
are still affected with certain weak points. Optical systems
based on multi-line sensor technology (ADS40, HRSC) are able
to seriously compete with laser scanner systems in some
application areas. The generation of Digital Surface Models
(DSMs) from multi-line scanner data has been investigated
thoroughly (Scholten and Gwinner, 2003) and has reached a
certain level of practicability. Unfortunately these models are
so-called raster models; hence the depicted features, e.g.
buildings, are not classified and without topological
information.
Besides airborne acquisition techniques high-resolution
spaceborne scanners offer a very important data source, since
they can capture images from nearly any region of the earth's
surface without having any juridical problems.
This paper focuses on aspects concerning city model generation
from both airborne and spaceborne line-scanning systems,
while it does not deal with the derivation of DSMs from those
images. By analysing (already existing) DSMs and therefrom