obtain edges and nodes, a vectorization is necessary
(Fig. 12).
A)
=> jé
2 ss S qu Eb BRR mu
E mm B NS i qu = ERR m c | © i
NS S v m S m
i EHH
a us
X 5 d
i
Sn AS A
gU Nu = _ | ius S us
po b 2 E SR
| |
ü
as
Figure 12: Vectorized Edges and Nodes
The next step was the feature-based matching. For this
task we chose three images with tilt angles of - 5°, 0° and
+ 5°. As the result of the matching process over 90% of
the features, both nodes and edges, were matched suc-
cessfully. The matched nodes were used as tie points for
the following estimation of the orientation parameters.
After the evaluation of the three-dimensional coordinates
of the edge points, the reconstructed microsample has
been visualized, as seen in Fig. 13.
Figure 13: Automatically Reconstructed Silicon Sample
(Dimensions in um)
5. OUTLOOK
It is possible to analyse surfaces in microranges with the
described photogrammetric methods. This enables the
comparison of quantitative differences, for example be-
tween the nominal and the actual dimensions influenced
by the technological process.
Further improvements are necessary to develop the pre-
sented photogrammetric system to a robust automatic
method in three-dimensional microsciences. With help of
a detailed calibration of SEM it is possible, to simplify the
orientation process with the aim of a higher stability and
accuracy. In order to further automation the number of
threshold values and control parameters should be re-
duced. Other correspondence algorithms will be investi-
gated in the future, for instance relational matching
methods. In some cases, the difficult process of feature
matching can be supported by a knowledge-base from
evaluated data from the CAD-model of the microstructure.
ACKNOWLEDGEMENTS
The joint project ,Microtopography" between the Tech-
nical University of Berlin and the Institute for High Tech-
nology Jena is financially supported by the Deutsche
Forschungsgemeinschaft (DFG).
REFERENCES
Burkhardt, R., 1981. Die stereoskopische Ausmessung
elektronenmikroskopischer Bildpaare und ihre Genauig-
keit. Methodensammlung der Elektronenmikroskopie, 10,
Wissenschaftliche Verlagsgesellschaft Stuttgart, pp. 1-59.
Canny, J., 1986. A Computational Approach to Edge
Detection. IEEE Transactions on pattern analysis and
machine intelligence, Vol. PAMI-8, No. 6, pp. 679-698.
Ebner, H., Hofmann-Wellerhof, B., Reiss, P., Steidler, P.,
1980. HIFI — Ein Minicomputer-Programmsystem für
Hóheninterpolation mit Finiten Elementen. Zeitschrift für
Vermessungswesen, 105 (5), pp. 215-225.
El Ghazali, M. S., 1984. System Calibration of Scanning
Electron Microscopes. International Archives of Photo-
grammetry and Remote Sensing, Rio de Janeiro, Brazil,
Vol. XXV, Part A5, pp. 258-266.
Fórstner, W., 1986. A feature-based correspondence al-
gorithm for image matching. International Archives of
Photogrammetry and Remote Sensing, Rovaniemi, Fin-
land, Vol. XXXVI, Part 3/8, pp. 150-166.
Gleichmann, A., Kóhler, J. M., Hemmleb, M., Albertz, J.,
1994. Photogrammetric Determination of Topography of
Microstructures by Scanning Electron Microscope. San
Jose, CA, USA, SPIE Proceedings, Vol. 2184, pp. 254-
263.
Hemmleb, M., Albertz, J., Schubert, M., Gleichmann, A.,
Kóhler, J. M., 1995. Photogrammetrische Bestimmung der
Krümmung einer Mikrokantilever-Probe mittels Raster-
elektronenmikroskop. Beitráge zur elektronenmikroskopi-
schen Direktabbildung und Analyse von Oberflächen, Vol.
28, pp. 65-72.
König, G., Nickel, W., Storl, J., Meyer, D., Stange, J.,
1987. Digital Stereophotogrammetry for Processing SEM
Data. Scanning, 9 (5), pp. 185-193.
Li, R., 1990. Erfassung unstetiger Oberflächen aus digi-
talen Bilddaten durch Flächen- und Kantenzuordnung.
Deutsche Geodätische Komm., R. C, No. 364, Munich.
Rüger, W., et al., 1987. Photogrammetrie. 5th edition.
VEB Verlag für Bauwesen, Berlin, pp. 20-21.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
A NOVEL S
KEY WORD
ABSTRACT
At the same :
obtained usin,
algorithm, hu
wave-field wa
measurement:
Pe
Figure 1: Ex;
facility (perim
Institute for E
(Germany).
In this paper
bines two sp
slope imagin
already been
put together
vides all rel:
cal wave nu
a function c
[Balschbach
