anbul 2004
Adobe
1 Road,
039-7900.
03. Inves-
10te Sens-
XIV, Part
ebäudere-
llen und
t, Institut
nmission,
(accessed
> von Ob-
Deutsche
sification
York.
by Com-
ch Kom-
sis, Uni-
Deutsche
| Analyse
Iberg.
Vergleich
ichender
ngsnach-
ified 3D
. IV Joint
, Integra-
accessed
he scope
s for the
complex,
inded by
THREE-DIMENSIONAL MODELLING OF BREAKLINES FROM AIRBORNE
LASER SCANNER DATA
Christian Briese
Institute of Photogrammetry and Remote Sensing
Vienna University of Technology, Guflhausstrafie 27-29. A-1040 Vienna, Austria
cb@ipf.tuwien.ac.at
Commission IIT, Working Group III/3
KEY WORDS: Laser scanning, LIDAR, DEM/DTM, Three-dimensional, Breakline, Mod
ABSTRACT
Airborne laserscanning allows a very detailed sampling of the 1
procedure. For the representation of models computed on the basis of this dat
uetworks (TINs) are used, which do only implicitly store breakline information. For high quality surf
breaklines must be explicitly stored within the data structure. Therefore a 3D vector representation of the break
necessary. This paper presents a method for the modelling of 3D breaklines not only from airborne las
but from any kind of point cloud data. The method is based on a pairwise intersection of rol
elements along the breakline. It allows a modelling on the basis of original irregular distributed point cloud
elling.
andscape within a more or less automated recording
a mostly raster or triangulated irregular
ace modelling
lines is
er scanner data
oustly estimated surface
S even in
wooded areas. For this procedure a 2D approximation of the breakline is required. Therefore one section concentrates
on the determination of these initial values in order to automate the process. Results of the modelling are presented in
an examples section. À summary with an outlook on future developments concludes the paper.
1 INTRODUCTION
Airborne laserscanning (ALS, also referred as LIDAR) al-
lows a very dense sampling of the landscape with the help
of high frequent range and angle observations. Additional
synchronised measurements of the position and orientation
of the scanner unit within a global co-ordinate system al-
low the computation of a geo-referenced point cloud. On
the basis of this data and by the use of algorithms with a
knowledge base for a specific application a lot of different
models (e.g. terrain, building or vegetation models) can be
determined (cf. (Axelsson 2000), (Brenner 2000), (Wack
et al. 2003)).
For the representation of the surface models generated
from ALS data or other automated data acquisition meth-
ods (e.g. image matching techniques) mostly raster resp.
grid models or triangulated irregular networks (TINs) are
in use. In general these models are only computed on the
basis of an irregular distributed point cloud and therefore
they do only implicitly store breakline information. The
quality of the breakline description within these models de-
pends next to the original point sampling interval on the
size of the stored raster resp. triangle cells. In contrast to
these models without an explicit breakline description, it is
essential for high quality surface models (e.g. for hydrolog-
ical applications) to store breakline information explicitly
in the data structure. For this aim a three-dimensional
vector representation of the breaklines is necessary. Based
on this description an integration of relevant breaklines
into a constrained triangulation process or into a hybrid
raster data structure ( (Kraus 2000), cf. figure 1) can be
performed. Additionally, breaklines are very important for
the task of data reduction. They help to describe surface
discontinuities even in models with big raster resp. trian-
gele cells.
1097
Figure 1: Perspective view of a hybrid raster digital ter-
rain model including breakline information.
Next to the fundamental role of breaklines for the final
surface models the explicit description of these disconti-
nuity lines is very important for the generation of digital
terrain models (D'TMs) from ALS data. For the determi-
nation of a D'TM on this data basis a classification (filter-
ing) of the acquired point cloud into terrain and off-terrain
points is required. For this aim a lot of different algorithms
were developed (e.g. (Axelsson 2000), (Kraus and Pfeifer
1998), (Vosselman 2000)). An international filtertest (Sit-
hole and Vosselman 2003) showed deficits of all methods
used by the participants in the areas of surface disconti-
nuities even when the algorithm has some special rules to
avoid misclassification next to breaklines. To cope with
this problem an integration of explicit modelled breaklines
during the whole DTM generation procedure is essential.
However, this makes a modelling of breaklines, based on
the original unclassified ALS points, necessary.
After a short summary of the current state of research in
the area of breakline modelling from ALS data the paper
