289
ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001
finding altimetric accuracy. Based on the same data set, the
result can be shown in Table 3.
Table.3 Result of experimented data considering at the
relationship model of x.y.v.u
Calculated
value - E
CT point No.
Number of
CT
Combinations
0.913006
0123456
7
1
0.993801
0123467
8
8
0993801
0123467
9
36
0.994277
0124589
10
116
0.996060
O 1 2467 10
11
312
0.996060
0 1 2467 10
12
730
0.997054
0 1 348 10 12
13
1547
0.997625
1 3458 12 13
14
3044
0.997625
1 3458 12 13
15
5620
0.997625
1 34581213
16
9850
0.998027
1 34 10 11 13 16
17
16543
0.998323
1 3 10 11 13 16 17
18
26819
0.998325
1 3 10 11 13 16 18
19
42172
0.998325
1 3 10 11 13 16 18
20
64607
0.998325
1 31011 13 1618
21
96678
0.998325
1 3 10 11 13 16 18
22
141617
0.998325
1 3 10 11 13 16 18
23
203365
0.998328
1 3 11 13 16 1823
24
287262
situation, conditioned by the way to made extraction, the highest
altimetric and planimetric accuracy were 0.998480 (E) and
0.998328 (E) respectively. Therefore, It can be briefly concluded
that (a) polynomial function can represent 3D building objects
from a 2D photograph without the shadow of photogrammetric
(b) accuracy of 3D building objects in vertical and horizontal axis
can be acceptable (c) few control points can serve to get
acceptable result. These entire conclusions were advantage of
this system. On the other hand, the disadvantage of the system
was computational time for generating the best-fit solution of
data. It often made long period of time, when it had to work on
huge data.
For further work, an automated system of extraction will replace
the manual system, in order to improve the accuracy. Also the
3D-building-objects visualization will be available.
ACKNOWLEDGEMENTS
The authors are very grateful to Justin Hickey, Dr.Vivarad and
Dr.Duan for some advices and also Daniel Kassahun for
sequential idea and English proofing.
The E value was over 0.99, when it was 8 control points. As
same as Table 2, the E value increased, while the number of
control points was increased, together with number of
combinations. Finally, It was a line graph (as in Figure 4),
representing the changing of E value.
According to this graph, it was eight control points that was
enough for generating 3D building objects particularly in
planimetric form. E value was initially above 0.99,and then went
on in this level until the last control points.
REFERENCES
[1] Brunn, A., Guelch, E., Lang, F., Foerstner, W., 1998. Hybrid
Concept for 3D Building Acquisition. ISPRS Journal of
Photogrammetry and Remote Sensing v 53, 119-129.
[2] Kanade, T., et al., 1998. Constructing Virtual Worlds Using
Dense Stereo. Proceedings of sixth IEEE International
Conference on Computer Vision, 3-10.
[3] Peuquet, D.J., 1984. A Conceptual Framework and
Comparison of Spatial Data Models. Cartographica v 21, no.4,
66-113
[4] Stigant, S.A., 1959, The elements of determinants, matrices
and tensors for engineers. Macdonald, London.
[5] Van den Heuvel Frank A, 1998. 3D reconstruction from a
single image using geometric constraints, Isprs Journal Of
Photogrammetry And Remote Sensing (53)6 pp. 354-368
BIOGRAPHY
Taravudh Tipdecho:
1) Doctoral Candidate, Space Technology Applications and
Research (STAR) Program, School of Advanced Technologies
Asian Institute of Technology, THAILAND
2) Researcher, High Performance Computing Laboratory
(HPCC), National Electronics and Computer Technology Center
(NECTEC), Bangkok, THAILAND.
Xiaoyong Chen:
1) Associated Professor, Space Technology Applications and
Research (STAR) Program, School of Advanced Technologies
Asian Institute of Technology, THAILAND
Figure 4. (a) the deviation of E value of relationship model of
x,y,v,u (b) the corresponding relative changes of E value.
5.CONCLUSIONS
Based on the model of relationship represented in polynomial
form, eight control points in horizontal (x,y) and ten control
points in vertical (height(z)) had sufficient quality to generate the
3D building objects for this study area. For specific tested