144
ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
Fig.3 Distribution Fig.4 Fiducials (a)
(b)
(c)
(d) Fig.5 Similar Object
Table 3 the candidates of each fiducial mark
Fiducial
X (pixel)
Y (pixel)
Correlation
coefficient
1
1
2935
205
0.47895
2
2935
169
0.42382
3
2886
205
0.41220
2
4
67
1598
0.55935
5
68
1655,
0.43236
6
67
1550
0.41243
3
7
2965
4453
1.00000
8
2965
4420
0.69096
9
2965
4508
0.69037
4
10
5817
1560
0.28147
11
5817
1609
0.22221
12
5918
1582
0.21994
Fig.4(a) is a good image of fiducial mark. Now one object similar
to the fiducial mark is added to generate the image in Fig.5. The
new image is used to test the algorithm. The candidates of
fiducial mark 3 are listed in Table 5. The point number of added
object is 8, whose correlation coefficient is also large. The
parameters of the algorithm are set the same as before. After
T
correct one-to-one correspondence is got.
Table 5 the candidates of the third fiducial mark
Fiducial
X (pixel)
Y (pixel)
Correlation
coefficient
3
7
2965
4453
0.99911
8
2913
4408
0.99513
9
2965
4408
0.70765
4 CONCLUSION
The interior orientation becomes much harder when fiducial
marks merge in image of objects. The self-adaptive algorithm
performs quite well without the usage of correlation coefficient.
ACKNOWLEDGE
This paper is supported by the National Natural Science
Foundation of China (Grand number: 40023004, 40001018)
Table 4 the correspondence matrix of real image experiment
m ai
1
2
3
4
5
6
7
8
9
10
11
12
Outlier
1
0.463
0.299
0.237
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
2
0.000
0.000
0.000
0.482
0.058
0.460
0.000
0.000
0.000
0.000
0.000
0.000
0.000
3
0.000
0.000
0.000
0.000
0.000
0.000
0.499
0.316
0.185
0.000
0.000
0.000
0.000
4
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.551
0.435
0.014
0.000
Outlier
0.537
0.071
0.763
0.518
0.942
0.540
0.501
0.684
0.815
0.499
0.565
0.986
Table 6 the correspondence matrix of the experiment with outlier
m a \
1
2
3
4
5
6
7
8
9
10
11
12
Outlier
1
0.376
0.298
0.326
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
2
0.000
0.000
0.000
0.408
0.184
0.407
0.000
0.000
0.000
0.000
0.000
0.000
0.000
3
0.000
0.000
0.000
0.000
0.000
0.000
0.393
0.242
0.365
0.000
0.000
0.000
0.000
4
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.502
0.372
0.126
0.000
Outlier
0.624
0.702
0.674
0.592
0.816
0.593
0.607
0.758
0.635
0.498
0.623
0.874
REFERENCES 1
(1) . Zhang Z., Zhang J.. 1996. Digital Photography. Press of
wtusm
(2) . Christian Heipke 1997 .Automation of interior , relative ,
and absolute orientation .ISPRS Journal of
Photogrammetry & Remote Sensing 52(1997)1-19
(3) . Kersten T and Haring T.1995 Automatic interior
orientation Internal report, ETH Zurich
(4) . Lue,Y, 1995. Fully operational automatic interior orientation
Proceedings Geolnformatics’95,HongKong,May
26-28,Vol-1,pp.26-35
(5) . Schcker.W. 1995. Ein operationeiles Verfahren zur
automatischen inneren Orientierung von
luftbildern .Z.Photogramm,fernerkundung,63(3):115-122
(6) . Strackbein,H.and enze.M.,1995.Automatische innere
Orientierung.Internal report .Landesvermessungsamt
Nordrhein Westfalen
(7) . S.Gold, A. Rangarajan, C. P. Lu, S Pappu, and E. Mjolsness.
New algorithms for 2-D and 3-D point matching : pose
estimation and correspondence . Pattern Recognition,
31 (8): 1019 1031, 1998
(8) . S.Gold, and A. Rangarajan. A graduated assignment
algorithm for graph matching. IEEE Trans. Patt. Anal.
Mach. Intell., 18(4):377-388,1996.
(9) . Haiti Chui and Anand Rangarajan 2000 A new algorithm for
non-rigid point matching , IEEE Conference on Computer
Vision and Pattern Recognition (CVPR),
(10) . Girosi, F.,Jones,M.,and Poggio,T.(1995). Regularization
theory and neural net work architectures. Neural
Computation, 7:219-269
(11) . Papadimitriou, C. and Rangarajan, K. (1982).
Combinatorial Optimization. Prentice-Hall,Inc.,Englewwood
Cliffd.NJ.
(12) . Gold,S.,Rangarajan,A.(1996). Softmax to softassign:
neural network algorithms for combinatorial optimizations
Journal of Artificial Neural Networks, in press
