ments, despite the inferiority of the images that were used for
the matching. These results point to the advantage of the
matching procedure on the manual measurements. Points
that appear on more than two images are matched simultan-
eously, while during the manual procedure they are measured
in pairs. The matching procedure therefore contributes to the
consistency in identifying exactly the same point on more than
two images. Overall accuracy of the matching was between
1/5 and 1/12 of a pixel.
5 SUMMARY
A method for accurate and reliable image matching of points
for aerotriangulation was developed and tested. In order to
increase reliability, large image patches that are more likely to
contain significant information are necessary. Most match-
ing methods use small image patches because the object
surface around the matching area is assumed planar. In
the method presented here, the matching is performed in
the object space which minimizes the geometric differences
between the matched patches. Consequently, much larger
image patches are used and the accuracy and the reliability
are increased.
The method has been derived theoretically and then imple-
mented in order to show its effects on aerotriangulation. The
experimental results described earlier are summarized in the
following determinations:
e The accuracy of the matching results is 1/12-1/5 of a
pixel. Although images of relatively coarse resolution
were used (30-60 um), the results compare favorably
with manual aerotriangulation. If higher resolution im-
ages were used for the matching, better results would
have been expected.
e The proposed matching scheme performs better than
traditional matching methods in areas that include sur-
face discontinuities. The results are similar to image-
space matching when smooth surfaces are involved.
e Since multiple images are simultaneously matched, the
method is superior to manual measurements because
a human operator cannot measure on more than two
images simultaneously.
These results are very encouraging. They show that by em-
ploying multiple-patch matching in the object space, conjug-
ate points for aerotriangulation with accuracy that exceeds
manual measurements may be obtained. Together with the
increase in efficiency by an automated procedure, these res-
ults lead to a new era of automated, accurate and reliable
digital aerotriangulation.
REFERENCES
Ackermann, F. and V. Tsingas. Automatic digital aerial
triangulation. In Proceedings, ACSM/ASPRS Annual Con-
vention, Vol. 1:ASPRS, pages 1-12, 1994. Reno, NV.
Ackermann, F. Automation of digital aerial triangulation. In
Proceedings, 2nd Course in Digital Photogrammetry, 1995.
Agouris, P. Multiple Image Multipoint Matching for Auto-
matic Aerotriangulation. PhD thesis, Department of Geo-
detic Science and Surveying, The Ohio State University,
Columbus, Ohio, 1992.
404
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
Baltsavias, E. P. Multiphoto geometrically constrained
matching. Mitteilungen 49, ETH-Zurich, 1991.
Dhond, U. R. and J. K. Aggarwal. Structure from stereo—a -
review. IEEE Transactions on Systems, Man, and Cybernet-
ics, 19(6):1489-1510, 1989.
Ebner, H., C. Heipke, and M. Holm. Global image match-
ing and surface reconstruction in object space using aerial
images. In Barrett, E. B. and D. M. McKeown, Jr., editors,
Integrating Photogrammetric Techniques with Scene Ana-
lysis and Machine Vision, pages 44—58, Orlando, FL, April
1993. Proceedings, SPIE 1944.
Helava, U. V. Digital comparator correlator system. /nter-
national Archives of Photogram metry and Remote Sensing,
27(B2):160-169, 1988.
Krupnik, A. Multiple-Patch Matching in the Object Space
for Aerotriangulation. PhD thesis, Department of Geodetic
Science and Surveying, The Ohio State University, Colum-
bus, Ohio, 1994.
Krupnik, A. On using theoretical intensity values as un-
knowns in multiple-patch least-squares matching. Submitted
to PE & RS, 1996.
Lemmens, M. J. P. M. A survey of stereo matching tech-
niques. International Archives of Photogrammetry and Re-
mote Sensing, 27(B8):V11-V23, 1988.
Schenk, T. and C. Toth. Towards a fully automated aerotri-
angulation system. In Proceedings, ACSM/ASPRS Annual
Convention, Vol. 3, pages 340-357, New Orleans, LA, 1993.
Toth, C. and A. Krupnik. Concept, implementation and res-
ults of an automatic aerotriangulation system. Photogram-
metric Engineering and Remote Sensing, 1996. In Press.
Wrobel, B. P. Least-squares methods for surface reconstruc-
tion from images. International Archives of Photogrammetry
and Remote Sensing, 27(B3):806-821, 1988.
KEY V
ABSTI
Autom
time c
triangt
full ad
The ps
point
systen
oriente
and bl
rigoroi
The pe
contro
as wel
which
autom
demotr
results
Digita
level |
impro
autorr
provic
proce
power
DEM
auton
have
digital
partic
it is C
digita
triang
in a d
be sic
Sever
scien
well.
Toni
(Sche
and tl
Tsing
DeVe
syste
appre
respe
in cc
autor
