International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
some difficulties were encountered and they are discussed
below.
Although the segmentation procedure shows successful results,
it might fail to segment roof regions in some areas. Areas where
the roof segment is not smooth or its size is not large enough to
contain enough LIDAR points to estimate reliable geometrical
parameters of the segment are some examples which might lead
to inaccurate roof segments. Significant existence of small
structures over a small roof region if added to the original noise
in LIDAR data may cause the production of noisy parameters
during the plane fitting procedure and consequently unreliable
segmented regions. However, in such cases, increasing the data
density might alleviate this obstacle to a certain extent. Another
example of segmentation failure occurs where adjacent trees are
extended over a large part of the roof facet that causes an
occlusion where not all laser pulses can reach the building roof.
This situation can be avoided by a good planning for the survey
time where there are no leaves which would minimize
occlusion.
In roof polygon extraction, the performance of the simple and
complex roof polygon extraction was successful especially with
large roof regions as shown in figures above. Roof polygons
were extracted and successfully connected. However, some
nodes might be shifted from their true position during the
joining and connecting of the roof planar segments especially
with complex buildings. On the other hand, the performance of
the planar roof connecting algorithm deteriorates in the
presence of very small close by roof regions. This is due to the
fact that polygon vertices may be so close to each other that
they incorrectly forced to coincide during the connection
procedure.
ACKMOWLEDGEMENT
The authors would like to acknowledge the support for this
research from the following organizations: the Army Research
Office and the Topographic Engineering Center.
REFERENCES
Alharthy, A., Bethel, J. 2002. Heuristic filtering and 3d feature
extraction from LIDAR data. /SPRS Commission lll,
September 9 - 13, 2002, Graz, Austria.
Brenner, C. 2000. Towards fully automated generation of city
models. /SPRS, vol. XXXIII, Amsterdam 2000.
Brenner, C., Haala, N. 1999. Extracting of building and trees in
urban environments. /SPRS J vol. 54, 130-137.
Brunn, A. 2001. Statistical interpretation of dem and image data
for building extraction. In A. Grün, editor, 4SCONA-Workshop
2001. Balkema-Verlag.
Brunn, A., Weidner, U. 1997. Extracting buildings from digital
surface models. JAPRS Vol. 32, part 3-4w2, stuttgart.
Mikhail, Edward M., Ackermann, F., 1976. Observation and
Least Square. IEP-A Dun-Donnelley Publisher, New York.
Vosselman, G. 1999. Building reconstruction using planar faces
in very high density heights data. /4PRS, vol. 32, part 3-2w5,
87-92.
Weidner, U., Forstner, W., 1995. Towards automatic building E?
extraction from high resolution digital models. /SPRS J. 50 (4),
34-49.
Divis.
KEY V
ABSTI
This p:
clusteri
on the
constiti
propert
for the
capture
Laser-s
acquisi
digital
widely
(Acker
acquire
from tr
circum:
spatial
the bar
various
an atter
derivati
get a D
non-ter
This pt
operatic
classifi
comple
filtering
assump
charact
terrain
For inst
the low
terrain |
than a ¢
with sn
DTM.
betweer
of indi
weights
Vosselr
momen
triangul
based fi
large he