bul 2004
growing.
etter than
Jaries are
hbouring
ce in the
Voronoi
00). It is
e planar
Voronoi
es.
Centre:
merging
m.
ortion of
planes is
1e, a step
split into
s can be
hbouring
; critical,
ta alone.
gons can
nages by
m image
as to be
ch. edges
tion that
Thus, in
roject all
roximate
ements /,
t parallel
'tors, e.g.
segments
n another
to be the
s straight
Its of the
:ombined
segment
two such
in object
t edges.
I»; edges
'ering the
hed with
is based
ely loose
roximate
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
polygon. Multiple matches are eliminated based on a minimum
distance criterion. The results for the building in figure 3 are
presented in figures 7 and 8. The shapes of the boundary
polygons could be improved considerably. However, the
polygons are not yet completely correct at the building corners.
This could be overcome either by instantiating hypotheses
about regularities in areas that do not receive support from
image features, or by iterating the matching technique in cases
where the approximations are not good enough. There is a small
displacement between the intersection lines and image edges,
which is either caused by errors in the geo-coding or in the
plane parameters. This emphasises the importance of a final
adjustment, taking into account both the LIDAR points, the
image edges (both for step edges and intersection lines), and
geometric constraints, using the adjustment model described in
(Rottensteiner and Briese, 2003) The considerable improvement
of the shapes of the building outlines as compared to figure 5
also improves the prospects for the success of such an
adjustment.
JT : C T t
t ÀÀ Ae à 7 7 =
fo a E 25. i een
: . —À > UT
Figure 7. Delineation of step edges for four roof planes.
Dashed lines: 3D edges. Dotted line: approximate
polygon. Full line: polygon after matching.
Figure 8. Left: roof polygons after matching. Right: back-
projected to one of the aerial images.
4. CONCLUSION AND FUTURE WORK
We have presented a method for building detection from
LIDAR data and multi-spectral images, and we have shown its
applicability in a test site of heterogeneous building shapes. The
method is based on the application of the Dempster-Shafer
theory for classification. The results achieved were very
satisfactory. The detection rate for buildings larger than 50 m?
was 95%, and about 89% of the detected buildings were correct.
The detection rates decrease considerably with the building
size: building structures smaller than 30 m^ could generally not
be detected. In this context, future work will concentrate on
evaluating the relative contribution of the cues used for
classification. We also want to extend the evaluation to the
influence of the LIDAR resolution on the results.
We have also shown how aerial images and LIDAR DSM can
be combined to improve both the results of roof plane detection
and the shapes of the roof boundaries. This is still work in
progress, and the algorithms involved can be improved in many
ways. For instance, moments or other invariants of the image
segments could be considered in matching, especially if new
planar segments are to be introduced based on evidence from
the images. The matching of 3D straight lines and roof polygon
segments could be expanded to include more robust techniques
for outlier detection. However, we have already shown some of
the benefits that can be achieved by using multiple data sources
for the reconstruction of buildings by polyhedral models.
ACKNOWLEDGEMENTS
This work was supported by the Australian Research Council
(ARC) under Discovery Project DP0344678 and Linkage
Project LP0230563. The LIDAR data were provided by
AAM Hatch, Mt Colah, NSW 2079, Australia
(www.aamhatch.com.au). The aerial images are courtesy of
AAM Hatch and Sinclair Knight Merz.
REFERENCES
Ameri, B., Fritsch, D., 2000. Automatic 3D building
reconstruction using plane roof structures. In: ASPRS congress,
Washington, DC, accessed in the WWW (27/01/2004):
www.ifp.uni-stuttgart.de/publications/2000/pub2000.html.
Brunn, A., Weidner, U., 1997. Extracting buildings from
digital surface models. In: JAPRS XXXII / 3-4W2, pp. 27-34.
Forstner, W., 1994. A framework for low level feature
extraction. In: Computer Vision - ECCV '94 Vol. II, Sm ICCV.
Boston, MA, pp. 383-394.
Haala, N., Brenner, C., 1999, Extraction of buildings and trees
in urban environments. /SPRS J. Ph & RS 54(2-3), pp. 130-137.
Heipke, C., Mayer, H., Wiedemann, C., Jamet, O., 1997.
Evaluation of automatic road extraction. In: /APRS XXXII /
3-2W3, pp. 47—56.
Klein, L., 1999. Sensor and data fusion, concepts and
applications. 2" edition, SPIE Optical Engineering Press.
Lu, Y.-H., Trinder, J., 2003. Data fusion applied to automatic
- building extraction in 3D reconstruction. In: Annual ASPRS
817
conference, Anchorage, Alaska, pp. 114-122.
Rottensteiner, F., Briese, C., 2003. Automatic generation of
building models from LIDAR data and the integration of aerial
images. In: /APRSIS XXXIV / 3-W13, pp. 174-180.
Rottensteiner, F., Trinder, J., Clode, S., Kubik, K., 2003.
Detecting buildings and roof segments by combining LIDAR
data and multispectral images. In: Image and Vision
Computing New Zealand (IVCNZ 2003), pp. 60-65.
Schenk, T., Csatho, B., 2002. Fusion of LIDAR data and aerial
imagery for a more complete surface description. In: ZAPRSIS
XXXIV / 3A, pp. 310-317.
Voegtle, T., Steinle, E., 2003. On the quality of object
classification and automated building modeling based on
laserscanning data. In: ZAPRSIS XXXIV /3W13, pp. 149-155.
Vosselman, G., Dijkman, S., 2001. 3D building model
reconstruction from point clouds and ground plans. In:
IAPRSIS XXXIV / 3WA, pp. 37-43.
