nts after mismatch
ost Benefit
ction function
6 103
> 44
1 44
orientation from the
with two evaluation
erator. As shown, the
vere less accurate than
e results of automatic
cel accuracy, which is
to be used as a means
hly accurate matching
ive orientation.
nefit Analytical
ction plotter
Oum
44 9
‚618 90.586
450 -2.393
). 175 150.341
647 0.604
210 -1.327
2.168 359.064
12.6 10.3
)4.5 4.5
S
yo evaluation functions
e matching problem,
ion. As shown, the
ovided reliable results
urban area images
repetitive pattern by
s between two straight
eme, locally consistent
isistent solution. From
of obtaining a solution
ermore, the 2-D tree
) the tree search and
iciently. Implementing
the modified forward
ching reach a solution
functions in relational
n performs little better
ie domain of the image
obtaining an initial
ith little rotation was
d that the distinct and
1a 1996
prominent relations such as node relations were a good way of
solving the initial approximation problem. Despite the limited
experiments in this study, the results confirm that relational
matching successfully deals with many problems in urban area
images.
This research, however, showed that there were some
problems in relational matching. First, the computation
expenses are too heavy to obtain the solution in a reasonable
amount of time, compared to existing matching methods. Since
relational matching is a combinatorial optimization process, the
computing complexity is inevitable. Second, it behaves ina
poor manner when there are no conjugate features in a set of
features. Finally, the relationships between primitives must be
well constrained one another in order to achieve a reliable
solution. A highly abstract contextual information is a key to
reaching a reliable solution.
From the experiments in this study, the following issues are
identified for future research.
e A feature extraction algorithm which detects and interacts
well with the physical boundaries must be explored. As
shown in this study, wrong matches are always acquired
when there are no corresponding feature primitives.
e The higher the abstraction of feature description, the better
and faster the relational matching reaches a solution. A
way of acquiring a high level of abstraction of feature
description prior to relational matching must be explored.
e Some attributes in the primitives and relations are
orientation variant. In order for the proposed relational
matching scheme to be implemented for more general
cases, those attributes should be replaced with orientation
invariant ones.
e In this study, two descriptions are realtionally matched.
The proposed relational matching scheme can be extended
to match more than two descriptions simultaneously so
that it could be utilized for multiple image matching
problems such as automatic aerotriangulation.
e The computation complexity of relational matching in the
form of a tree must be reduced by investigating other
existing search methods such as relaxation technique,
maximal clique and simulated annealing.
e The proposed relational matching scheme implemented in
this research has the potential of being used for updating
maps, object recognition and navigation. Further
investigation and developments are required in the near
future.
REFERENCES
Ballad, Dana H. and Brown, Christopher M. (1982)
Computer Vision. Prentice-Hall, Inc.
Boyer, K. L., Kak, A. C. (1988) Structural Stereopsis for 3-D
Vision. IEEE Transactions on Pattern Analysis and Machine
Intelligence, Vol. 10, No. 2, March, 144-166. ,
Burns, J. B., Hanson, A. R., and Riseman, E. M. (1986)
Extracting Straight Lines. [EEE Transactions on Pattern
Analysis and Machine Intelligence, Vol. 8, No. 4, July, 425-
455.
Cho, Woosug (1995) Relational Matching for Automatic
Orientation. Ph.D Dissertation, The Ohio State University,
Columbus, Ohio.
Doorn, B., Agouris, P., Al-Tahir, R., Stefanidis, A. and
Zilberstein, O. (1990) Digital Stereo Matching: In Perspective.
Technical Notes in Photogrammetry No. 10, Department of
Geodetic Science and Surveying, The Ohio State University.
Columbus, Ohio.
Fórstner, Wolfgang and Gülch, E. (1987) A Fast Operator for
Detection and Precise Location of Distinct Points, Corners and
Centers of Circular Features. ISPRS Intercommissiom
Workshop, Interlaken, June, 281-305.
Haralick, Robert M. and Shapiro, Linda G. (1992) Computer
and Robot Vision. Addison-Wesley Publishing Company, Inc.
Moravec, H.P. (1977) Toward Automatic Visual Obstacle
Avoidance, Proceedings of the 5th Joint Conference on
Artificial Intelligence, Cambridge, 584.
Schenk, T. (1992) Algorithms and Software Concepts for
Digital Photogrammetric Workstations. Report No. 418,
Department of Geodetic Science and Surveying, The Ohio
State University, Columbus, Ohio.
Shahin, Fayez. (1994) An Automatic Correspondence for
Digital Images Using Linear Features. Ph D Dissertation,
Department of Geodetic Science and Surveying, The Ohio
State University, Columbus, Ohio.
Tang, L. and Heipke, C. (1994) An Automatic Procedure for
the Relative Orientation of Aerial Images. Proceedings on
Integration, Automation, and Intelligence in Photogrammetry,
Remote Sensing and GIS. 3rd Liesmars Colloquium, Wuhan,
October, 49-58.
Tsingas, Vassilios (1994) A Graph-Theoretical Approach for
Multiple Feature Matching and Its Application on Digital Point
Transfer. Munich Commission Ill, Institute for
Photogrammetry, University of Stuttgart, 865-871, Stuttgart.
FRG.
Vosselman, George (1992) Relational Matching. Lecture
Notes in Computer Science 628, Springer-Verlag, Berlin
Heidelberg.
Zilberstein, Ofer (1992) Relational Matching for Stereopsis.
International Archives of Photogrammetry and Remote Sensing,
Commission III, Washington D.C., 711-719.
119
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
