CA TA. KJ Mw iv im
iv
iv
Uw
Ansgar Brunn
A STEP TOWARDS SEMANTIC-BASED BUILDING RECONSTRUCTION USING
MARKOV-RANDOM-FIELDS
Ansgar Brunn
Institute for Photogrammetry, Bonn university, Germany
Brunn Q ipb.uni-bonn.de
KEY WORDS: Building reconstruction, Digital surface models, Markov-Random-Fields, Topology
ABSTRACT
In this paper we describe a new concept for the reconstruction of buildings. In contrast to most of the published ap-
proaches, we link the reconstruction process with the building interpretation. With this linkage we want to enhance the
reconstruction result and to yield semantic information about the buildings. We introduce building models based on their
topology. We also may use data from different sensor types. The analysis is done locally using statistical building infor-
mation for the interpretation in a Markov-Random-Field and using e. g. geometric or radiometric “appearance” models
for the reconstruction. A real data example from laserscanner observations demonstrates the approach.
KURZFASSUNG
In diesem Artikel beschreiben wir ein neues Verfahren zur Rekonstruktion von Gebáuden. Im Gegensatz zu den meisten
in der Literatur bereits veróffentlichen Verfahren, verbinden wir die Rekonstruktion mit der Interpretation. Dadurch
verbessern wir das Ergebnis der Rekonstruktion und erhalten zusätzlich semantische Information über das Gebäude. Wir
verwenden ein Gebäudemodell, das auf der Topologie der Gebäude definiert ist. Auserdem ist die Integration von Daten
unterschiedlicher Sensortypen möglich. Die Analyse der Daten erfolgt bei der Interpretation mit lokalem statistischen
Gebäudewissen und bei der Rekonstruktion mit lokalen Modellen der *Erscheinungsform" der Gebàude (z. B. geometrisch
oder radiometrisch). Ein Beispiel mit realen Entfernungsdaten demonstriert den Ansatz.
1 INTRODUCTION
In this paper a new concept for the reconstruction and interpretation of building data is introduced. We present an approach
for multi sensorial building data analysis which combines reconstruction and interpretation.
1.1 Motivation
Since a few years a large demand for urban and suburban 3D data can be recognized. Various applications need 3D data for
planning. Others need 3D data as background information for visualization and analysis. There is an increasing number
of applications (from GIS) who ask for interpreted data, which allow the application to distinguish between important and
non-important parts of urban data.
Although some approaches for building reconstruction have been presented in the last years, only a few approaches are
able to use different sensor types, mainly as different parts of a specific work flow, e. g. Haala and Brenner (Haala and
Brenner, 1997) starting from map data adding laser-scanner data they reconstruct buildings. Others just use data of one
sensor type (Brunn and Weidner, 1997, Haala and Brenner, 1997, Vosselmann, 1999, Baillard et al., 1999, Moons et al.,
1998). Fischer et. al. (Fischer et al., 1999) have described on a conceptual level a tower of feasible algorithms which
could be used for the reconstruction of buildings in general. Multi sensorial reconstruction and interpretation is important,
because different data from different sensor types can support the reconstruction result from different aspects. Only with
sensor fusion it becomes possible to use the new developed aerial sensors like laser-scanners, digital cameras or three-
lines-cameras together. In this paper, an algorithm for the use of different sensor types is described. All sensor types are
handled in an equal manner, no one dominates the algorithm.
In most of the published algorithms for building reconstruction interpretation is done only as a side step. The interpretation
is mostly done by classifying the type of geometric structure of the reconstruction. Mostly information of classifications
of the neighborhood is not used. Except Lang (Lang, 1999) achieves some semantic interpretation, considering geometric
classifications from the neighborhood.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 117
