International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Table.3 Results for 10 000 Photograph Scale
m Mean Emax Emin
(m) (m) (m) (m)
Analytically
Instrument
Photographs +0.53 -0.36 +3.45 -2.77
scale 1/10000
Progressive
sampling
Analytically
Instrument
Photographs +0.47 -0.32 +1.96 -1.57
scale 1/10000
Selective
sampling
Analytically
Instrument
Photographs +0.32 -0.13 +0.89 -0.97
scale 1/10000
Composite
sampling
Table.4 Results for 16 000 Photograph Scale
m, Mean Emax Emin
(m) (m) (m) (m)
PHOTOMOD
Software
Photographs | £0.96 -0.44 +1.65 -2.92
scale 1/16000
Directly
EComputer
Imaging
Photographs
scale 1/16000
10 m grid
interval
+0.99 0.19 +3.97 -2.67
In Table 4, it can be seen that the RMS errors obtained by
Photomod and eComputer Imaging are approximately same.
The values of the DEM derived from photogrammetrically
different methods can be expressed in terms of a fraction of
Contour Interval (CI). The results obtained from 1/10 000
photographs scale are approximately within a range from CI/4
to CI/6. Moreover, The results obtained from 1/16 000
photographs scale are approximately CI/5. These values are
already equivalent values, which are within a range from CI/3
to CI/5, obtained by Li 1994. Data used by Li 1994 were
obtained from contour data sets. The test areas used in Li 1994
are three of those used for the ISPRS DTM test which was
conducted by Working Group 3 of Commission III.
If values in Table 3 and Table 4 are expressed in terms of "per
mil of h", then these values (per mil of h) are from 0.20 to 0.34
(Table 5).
Table.5 DEM accuracy in terms of *per mil of
h"
Progres. | Selective ~~ Compo. Ecomp.
; ; ; Photomod
Sampling | Sampling | Sampling eromac Imaging
El mz/h (96 0) 0,34 0,3 0,21 0,2 0,2
4. CONCLUSIONS
Following conclusions can be made as a result of this study. At
restitution with analytic instrument, the most accurate method is
composite sampling according to RMS errors. Its RMS
improves 32% than those of selective method, %40 than those
of progressive method. Despite the fact that the selective
sampling method should deliver better results than the
progressive sampling, the results show that both methods give
approximately same RMS values. This shows the importance of
the operator’s role in the selection of points. The operator
might have measured few points than required in the flat parts
of the test area in this study. This drawback is removed by
adding grid points in the composite sampling method. The
results obtained by Photomod and eComputer Imaging software
are approximately the same. But, eComputer Imaging software
has a time-consuming problem. When the accuracies of the
DEM's derived photogrammetrically are compared with each
other in terms of per mille of flying height, the results obtained
by Photomod, eComputer Imaging software and composite
sampling method do not differ considerably. In this study, the
role of grid intervals was not investigated. It is evident that the
grid based methods give different results at different grid
intervals.
ACKNOWLEDGEMENTS
This paper is partly based on Ph.D. thesis and Master thesis
carried out by H. Karabork and prepared in. Selcuk University.
The authors express their gratitude to Kamil Karatas for
cooperation and Scientific Search Centre of Selcuk University
for financial support.
5. REFERENCE
ETHZ 1999, Closed Project in Institute of Geodesy and
Photogrammetry Swiss Federal Institute of Technology,
“Automatic Digital Terrain Model Generation”,
http://www.photogrammetry.ethz.ch/research/autodtm/p1.html
(accessed 21 April, 2004)
Gasior, D. 1996, Automatic Derivation of a DTM with the
Helava System, OEEPE Workshop On The
International /
Application
Lausanne
Gruen, À.
Symposium or
[Imagine Ortho
Karabork, 20
Otomatik Yo
Uzerine Bir
Natural and /
Konya
Karabork, H
Yükseklik M
Ilerlemeli Orn
Thesis, Gradu
Selcuk Univer:
Lee, S.M. 199'
Digital Elevat
with Digital
Geography Gr:
Li, Z. 1994, A
Terrain Model:
Journal of Pho:
Photomod Use