tional relative orientation parameters can be obtained if the
interior orientation parameters are known. Table 2 shows
several important parameters of the example, where p,
Table 2: automatic relative Orientation parameters
original structural | recognized | point MS | total
image size | match. level points error time
1444x1975 | 180x246 26 0.074pixel | 7' 22"
Bx Bz 9, (8) GNE OREL
-0.086 -0.345 -0.001 -9.756 -0.063
and B, are the ratios of the base components in X and Z
direction to Y direction. The total time of the whole compu-
tation from the pyramid generation till the calculation of the
relative orientation parameters on the original image level
takes about seven minutes by a PC486/33MHz.
3.2 Automated Aerotriangulation
Since a few years automated aerotriangulation are under
investigation [Schenk/Toth, 1993; Tsingas, 1991]. The
main work to be automated is the tie points transfer and
measurement. The conventional methods in this area have
difficulties for the automatic block preparation, i.g. to deter-
mine connections, overlaps and some initial matched
points for each model in the whole image block, if there is
no approximate values available. The structural matching
is just the right method to solve this problem. The example
shows that with the structural matching the tie points can
be recognized fully automatically in the whole image block
without any a priori information.
Fig. 6(a) shows a image block with 3x2 images. An opera-
tor has digitized the images simply according to the image
numbers. So two images from the different stripes have a
rotation about 180'. In addition there is also no consider-
ation about the flying direction. Therefore the photographic
centers in the same stripe have a up-down configuration
instead of usually left-right. This complicated case is not
rare, if the operator is not highly qualified in photogramme-
try or if the digital images come directly from digital scan-
ners. An automated photogrammetric system should be
able to deal with this case too. The conventional methods
are not suitable for it. Only with the structural matching the
tie points can be recognized fully automatically. The struc-
tural matching is used for each two images. Fig. 6(b) dis-
plays the recognized lines and points by the structural
matching. A large number of tie points can be obtained on
the original image level from the transfer and densification
of these recognized points and lines on the reduced
images. the accuracy of matched image points is around
0.1 pixel in the example.
4. CONCLUSIONS
The "black box" philosophy for photogrammetric opera-
tions, which means the automation in photogrammetry,
has been predicted by some photogrammetric experts
[e.g. Achermann, 1991]. With the application of the struc-
tural matching this philosophy is further realized. In this
contribution the fully automatic recognition of the corre-
sponding image objects is realized without to know any a
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
priori information except the image data, even though for
the irregular or non-metric images. Many photogrammetric
tasks can reach their highest automation level by means of
. the structural matching. The paper has shown that the rel-
ative image orientation, the tie point transfer and measure-
ment for the automated aerotriangulation and the other
image matching tasks can be fully automated with the
application of the developed methods and program system
for the structural matching. The time for autonomous pho-
togrammetry is coming.
REFERENCES
Ackermann, F., 1991, Structural changes in photogramme-
try, 43th Photogrammetric Week, Stuttgart.
Fórstner, W., Gülch, E., 1987, A fast operator for detection
and precise location of distinct points, corners and centers
of circular features, Proc. of Fast Processing of Photo-
grammetric Data, Interlaken, Switzland.
Greenfeld, J.S., Schulte-Hinsken, S., Müller, W., 1991, A
strategy for automated stereo model orientation using a
feature-based matching procedure, Proc. of ACSM-
ASPRS Annual Convention, 1991.
Haralick, R.M., Shapiro, L.G., 1993, Computer and Robot
Vision, volume Il, Addison-Wesley Publishing Company.
Lemmens, M.J.P.M., 1988, A survey on stereo matching
techniques, International Archives of Photogrammetry and
Remote Sensing (IAPRS), Vol.27B3, pp11-23.
Li, M., 1988, High precision relative orientation using fea-
ture based techniques, IAPRS, Vol.27B3.
Nilsson, N.S., 1982, principles of artificial intelligence,
Springer-verlag.
Schenk, T., et al, 1990, Zur automatischen Orientierung
von digitalen Bildpaaren, Zeitschrift für Photogrammetrie
und Fernerkundung (ZPF).
Schenk, T., Toth, C., 1993, Towards a fully automated
aerotriangulation system, Proc. of ACSM/ASPRS.
Tang, L., Heipke, C., 1993, An approach for automatic rel-
ative orientation, Optical 3D Measurement techniques ll,
Wichmann.
Tsingas, V., 1991, Automatische Aerotriangulation, Proc.
of 43. Photogrammetric Week, Stuttgart.
Vosselman, G., Haala, N., 1992, Erkennung topographis-
cher Pafipunkte durch relationale Zuordnung, ZPF, No.6.
Vosselman, G., 1994, Use of tree search methods in digital
photogrammetry, IAPRS, Vol.30, part 3/2.
Wang, Y., 1992, Image segmentation based on parameter
estimation, IAPRS, Vol.29B5.
Wang, Y., 1994, Strukturzuordnung zur automatischen
Oberflächenrekonstruktion, Dissertationen der Fachrich-
tung Vermessungswesen der Universität Hannover, No.
207.
Wang, Y., 1995, Ein neues Verfahren zur automatischen
gegenseitigen Orientierung der digitalen Bilder, ZPF.
ee
