1231
The International Archives of the Photogrammetiy, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
If the coordinate differences were bigger than 3a (% 99.73),
they were identified as gross errors and re-measured or
eliminated. The results are shown in the tables (Table 2-4).
Image
No. Of
GCP’s
DEM
No. Of
CP’s
RMSE (m)
X
У
xy
Aerial
Photo
11
Photog.
DEM
23
0.94
0.95
1.33
Aerial
Photo
11
Cartog.
DEM
22
1.20
0.98
1.55
Table 2. Differences on aerial photographs
Image
No. Of
GCP’s
DEM
No. Of
Check
Points
RMSE (m)
X
У
xy
Quickbird
4
Photog.
DEM
25
1.05
0.85
1.35
Quickbird
5
Photog.
DEM
25
1.04
0.84
1.33
Quickbird
9
Photog.
DEM
26
1.29
1.01
1.63
Quickbird
9
Cartog.
DEM
26
1.56
1.02
1.86
Quickbird
Orient.P
ara.
Cartog.
DEM
27
38.97
14.89
41.72
Table 3. Differences on Quickbird images
Image
No. Of
GCP’s
DEM
No. Of
Check
Points
RMSE (m)
X
У
xy
IKONOS
4
Photog.
DEM
25
2.25
1.54
2.73
IKONOS
4
Ikonos
DEM
25
2.25
1.43
2.67
IKONOS
5
Photog.
DEM
25
2.16
1.49
2.62
IKONOS
7
Photog.
DEM
26
1.28
1.55
2.01
IKONOS
7
Ikonos
DEM
26
1.28
1.54
2.00
IKONOS
7
Cartog.
DEM
26
1.37
2.19
2.58
IKONOS
Orient.
Para.
Ikonos
DEM
27
11.97
11.84
16.84
IKONOS
Orient.
Para.
Cartog.
DEM
26
11.58
10.66
15.74
IKONOS
7
STEREO
24
0.77
0.89
1.17
RMSE (z) = 0.94
Table 4. Differences on IKONOS images
3.3 Geometric Accuracy Assessment Results
The geometric accuracy results show generally that the high
resolution satellite images can be used in mid / large scale
(1:6.000 to 1:10.000) mapping processes. And we can say that;
■ Quickbird ortho-images with 9 GCP’s have better
accuracy than IKONOS ortho-images with 7 GCP’s. And,
Quickbird ortho-images having 4/5 GCP’s have an accuracy
that is very close to aerial photographs (Atak and Altan,
2006). Moreover it has been detected that the accuracy is
getting worse in IKONOS and getting better in Quickbird ortho
images if the number of GCP’s decreases. The reason of this is
the GCP’s quality. Likewise the studies using SPOT HRS data
by Reinartz et al. (2006) and CARTOSAT-1 data by Michalis
and Dowman (2006) imply that the same accuracy can be
reached with 3/4 GCP’s as using 28/39 GCP’s.
■ Using direct sensor orientation parameters given by the
companies (in 2002), IKONOS ortho-images have better
accuracy than Quickbird ortho-images. But today, it is possible
to reach better accuracy with new orientation parameters. In
addition, systematic errors have been observed in the
easting/north easting (across track) direction. In most studies it
has been noted that the RMS geopositioning accuracy is just
below !4 pixel in the cross-track direction and close to Vi pixel
in both the along-track direction and in height (Hanley and
Fraser, 2004). And a significant portion of the easting errors
may have been due to variations in satellite elevation,
especially when the IKONOS satellite imaged at low elevation
angles (Helder et al., 2003; Yamakawa and Fraser, 2004).
■ The results obtained from orthophotos using IKONOS-
DEM and photogrammetric DEM are very close. Therefore
IKONOS-DEM can be used instead of photogrammetric DEM.
■ After evaluating of the gross error positions, it is
determined that they are mostly (25 of 27 - % 92.6) located on
the edge / outside of GGP network or on rough area where the
height differences are very high. In the same way Bouillon et al.
(2006) determine that the errors are getting bigger if the slope
increases (Figure 2).
Figure 2. The check point which has the biggest error
■ Finally, the results show that the stereo IKONOS image
has the best accuracy and stereo IKONOS image allows
mapping processes up to 1:6.000 scale. On the other hand, it is
possible to realize mapping processes up to 1:7.000 - 1:7.500
scale from mono satellite ortho-images.
4. FEATURE COMPILATION ASSESSMENT
4.1 Mapping Areas and Map Productions
Three regions which have different characteristics have been
selected in the study area for feature compilation analyses.
Some criteria were taken into consideration in selection of
compilation areas (Atak and Altan, 2006). First; the areas have