nding estimated
2
on the data set
compared in de-
ta set which was
station. The ac-
xpected to be in
ment and defini-
ticipants can be
ings were repre-
' roofs, i.e. each
e comparison is
-coordinates be-
vidual reports of
f the differences
buildings!:
2 O7 TI
the range of 0.4
1e expected val-
ncertainty of the
) of methods us-
lely the DEM.
drawn from the
on can be solved
data alone - pro-
compared to the
ange data is very
hen objects dis-
en if not, generic
f object recogni-
sist of nearly pla-
building model to
> Reconstruction requires more detailed information both
on the data but also on the model side.
> All participants restricted themselves to simple building
models (parametric or prismatic) except for Fayek. The
generality of his model however is not strong enough
with respect to the weak range data. The approach
seems very promising when more detailed information
is available (higher sampling rate of range data).
> A combination of all available information proves to de-
liver very reliable results (cf. Haala). In the given data
set the DEM was derived by a matching process from
the stereo images, thus the DEM is certainly less ac-
curate that the original image information. Furthermore
the sampling rate in the DEM is lower than the pixel
size in the images. There are approaches to introduce
building hypothesis into the matching process in order
to derive a more accurate DEM (cf. [Maitre & Luo 1992]
or [Kim & Muller 1995]).
4 Conclusion
The integration of prior information, especially GIS-data has
not been exploited to the extent possible and epected.
The only approach relying on prior GIS-information is of
Lócherbach. He restricts himself to 2D-objects, although in
principle the object-model can be given in 3D. This issue is
of particular importance especially for the revision of existing
databases (e.g. national databases like ATKIS).
For the detection and reconstruction of buildings using DEM
in combination with stereo seems to be a very promising way
(see also [Collins, Hanson, Riseman & Schultz 1995]).
In summary, the integrated interpretation of data of different
data sources is only in the beginning. To date, still many re-
searchers rely on a single data source, which is analyzed with
specific strategies. This might be due to the fact, that this
variety of different information sources has not been avail-
able until recently, and many people are not yet aware of it.
With the availability of new sensors however (e.g. laser scan-
ners) and also the direct availability of high resolution image
data (e.g. airborne or space borne digital line scanners with
resolutions from some decimeter to some m) many diverse
information sources will be ready for use in the near future.
The importance of data fusion has also been clearly pointed
out at the ISPRS workshop in Stuttgart in November 1995
[Fritsch, Sester & Hahn 1995]. Especially for the extraction of
man-made objects great profits can be expected.
References
Collins, R., Hanson, A., Riseman, E. & Schultz, H. [1995],
Automatic extraction of buildings and terrain from aerial
images, in Grün, Kübler & Agouris [1995], pp. 169-178.
Fayek, R. & Wong, A. [1994], Triangular mesh model for nat-
ural terrain, in ‘Proc. of the SPIE Intelligent Robots and
Computer Vision XIII’, Boston.
Fierens, F. & Rosin, P. [1994], Filtering remote sensing data
in the spatial and feature domains, in ‘Conf. on Image
and Signal processing for Remote Sensing’, Vol. 2315,
SPIE, pp. 472-482.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
Fritsch, D., Sester, M. & Hahn, M. [1995], Minutes of the joint
workshop on integrated acquisition and interpretation of
photogrammetric data, Technical Report 1995/2, Institut
für Photogrammetrie, Stuttgart.
Fritsch, D., Sester, M. & Schenk, T. [1994], Test on image
understanding, in H. Ebner, C. Heipke & K. Eder, eds,
"Spatial Information from Digital Photogrammetry and
Computer Vision’, Vol. 30/3, ISPRS, Munich, Germany,
pp. 243-248.
Grün, A., Kübler, O. & Agouris, P., eds [1995], Automatic Ex-
traction of Man-Made Objects from Aerial and Space Im-
ages, Birkhäuser Verlag, Basel, Boston, Berlin.
Haala, N. [1995], 3D building reconstruction using linear
edge segments, in D. Fritsch & D. Hobbie, eds, ‘Pho-
togrammetric Week '95', Herbert Wichmann Verlag, Hei-
delberg, pp. 19—28.
Haala, N. & Hahn, M. [1995], Data fusion for the detection
and reconstruction of buildings, in Grün et al. [1995],
pp. 211—220.
Kim, T. & Muller, J.-P. [1995], Building extraction and verfifi-
cation from spaceborne and aerial imagery using image
understanding fusion techniques, in Grün et al. [1995],
pp. 221-230.
Locherbach, T. [1994], Reconstruction of land-use units
for the integration of gis and remote sensing data, in
H. Ebner, C. Heipke & K. Eder, eds, ‘Spatial Informa-
tion from Digital Photogrammetry and Computer Vision’,
Vol. 30/3, ISPRS, Munich, Germany.
Lotti, J.-L. & Giraudon, G. [1994], Adaptive window algorithm
for aerial image stereo, in ‘12th International Confer-
ence on Pattern Recognition’, IAPR, IEEE, Jerusalem,
pp. 701—703.
Maitre, H. & Luo, W. [1992], 'Using models to improve stereo
reconstruction', /EEE Transactions on Pattern Analysis
and Machine Intelligence 14(2), 269—277.
Schutte, K. [1994], Knowledge Based Object Recognition of
Man-Made Objects, PhD thesis, University of Twente,
Enschede, Netherlands.
Stilla, U. [1995], 'Map-aided structural analysis of aerial im-
ages’, Journal of Photogrammetry and Remote Sensing.
Stilla, U., Michaelsen, E. & Lütjen, K. [1995], Structural 3d-
analysis of aerial images with a blackboard-based pro-
ductionsystem, in Grün et al. [1995].
Trinder, J. & Li, H. [1995], Semi-automatic feature extraction
by snakes, in Grün et al. [1995], pp. 95-104.
Weidner, U. [1995], Building extraction from digital elevation
models, Technical report, Institut für Photogrammetrie,
Bonn.
Weidner, U. & Férstner, W. [1995], ‘Towards automatic build-
ing extraction from high resolution digital elevation mod-
els’, ISPRS Journal 50(4), 38-49.
