unnecessary terms in the matching process. This
therefore would yield less accurate results. The
minimum standard deviation for ABM and 2SM are
approximately 0.19mm and 0.13mm respectively.
This shows that accuracy from the 2SM method is
better than the ABM.
4.3 Computation Time
Computation time (min:sec)
Window size ABM 2SM
9x9 00:26 00:21
55 X55 07:38 06:32
101 X 101 34:26 33:51
Table 1. Computation time for the ABM and SM methods
Table 1 compares the computation time of the ABM and
SM methods for the cylinder (95 points) at window sizes
9, 55 and 101 pixels. The stopping criteria used for the
least squares solution in both methods are :-
(a) rate of convergence (Mikhail & Ackermann, 1976),
which is 0.01 pixel
(b) the magnitude of the corrections to the unknowns,
which is 0.01 pixel
(c) the maximum number of iterations, which is 16
(d) the detection of unstable/weak normal equations by
the Singular Value Decomposition (SVD) method
(Griffiths & Hill, 1985).
From Table 1, it can be seen that, the computation time
taken for the SM method at any window size is less than
those of the ABM. This shows that the proposed
functional model has improved, thus, converges more
quickly even though it has 9 (as opposed to 8 for the
ABM) parameters to solve.
5. DISCUSSION AND CONCLUSIONS
it has been shown that, the internal precision obtained
from the 18M and 28M methods is much higher than
ABM for both the plate and the cylinder. This suggests
that the functional model has been improved to fit the
observations more closely.
The use of second order parameters in matching plane
surfaces and for smaller window sizes has been found to
give inaccurate results. This is probably due to
overparametrisation. Therefore, surface model should be
kept simple (only first order) if surfaces to be measured
are plane or working with smaller windows. The accuracy
of the matched coordinates from the 1SM and 2SM
methods was also shown to be higher than the ABM.
This experiment has shown that the conventional ABM
functional model has been improved through the use of a
surface model. This is reflected in the attained accuracy ,
convergence, and computation time. It has also shown
that suitable surface models should be used when dealing
with different surfaces. Comparisons with more complex
ABM methods, such as the geometrically constrained and
global matching would be of great interest.
560
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
REFERENCES
Ackermann, F., 1984. Digital Image Correlation :
Performance and Potential Application in
Photogrammetry. Photogrammetric Record, 11(64), pp.
429-439
Baltsavias, E.P., 1991. Multiphoto Geometrically
Constrained Matching. Ph.D. thesis, Institute of Geodesy
and Photogrammetry. Mitteilungen Nr. 49, 221p
Crippa, B., Forlani, G. & de Haan, A., 1993. Automatic
Deformation Measurement from Digital Images. Optical
3-D Measurement Techniques ll (ed. Gruen/Kahmen),
pp. 557-563
Foerstner, W., 1982 . On the Geometric Precision of
Digital Correlation. Proceedings of ISPRS Commission Ill
Symposium, Rovaneimi, Finland, June 7 - 11, 1982. In
IAP 24(3), pp. 176-189
Griffiths, P. & Hill, I.D., 1985. Applied Statistics
Algorithms. Ellis Horwood Ltd., England, 307p
Gruen, A.W., 1985. Adaptive Least Squares Correlation :
A Powerful Image Matching Technique. South African
Journal of Photogrammetry, Remote Sensing and
Cartography. 14(3), pp. 175-187
Gruen, A.W. & Baltsavias, E.P., 1987. High Precision
Image Matching for Digital Terrain Model Generation.
Photogrammetria 42(3), pp. 97-102
Hahn, M. & Brenner, C., 1995. Area Based Matching of
Colour Images. ISPRS Intercommission Workshop *From
Pixels to Sequences”, Zurich. 30(Part 5W1), pp. 227-234
Heipke, C., 1992. A Global Approach for Least-Squares
Image Matching and Surface Reconstruction in Object
Space. Photogrammetry Engineering and Remote
Sensing, 58(3), pp. 317-323
Mikhail. E.M. & Ackermann, F., 1976. Observations and
Least Squares. Harper & Row, New York, 497p
Pertl, A., 1985. Digital Image Correlation with an
Analytical Plotter. Photogrammetria, 40, pp. 9-19
Rosenholm, D., 1987a. Multi-Point Matching Using the
Least-Squares Technique for Evaluation of Three-
Dimensional Models. Photogrammetric Engineering and
Remote Sensing, 53(6), pp. 621-626
Rosenholm, D., 1987b. Least Squares Matching Method :
Some Experimental Results. Photogrammetric Record
12(70), pp. 493-512
Wrobel, B.P., 1991. Least-squares Methods for Surface
Reconstruction from Images. ISPRS Journal of
Photogrammetry and Remote Sensing, 46, pp. 67-84
KEY |
ABST
Conjug
polync
platfor
needec
summ:z
found
that the
matric«
no adv
model
seems
model
Algorit
spaceb
with ki
map pi
imager
control
minimi
develo
Most r
board 1
platforı
of the |
usually
for the
measur
with tii
effient
impro
parame
platforr
Other |
have be
the us
commc
