1271
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B4. Beijing 2008
than 0.4m in X, Y plane and 0.3m in height. The example
projects of the ground control points showed on the (Figure 1).
Two geometric correction variants have been used to correct
Resurs DK-1 images. The analysis of the influence of number
and distribution of the control points in boundary of the scene
on the result of geometric correction was realized in each
variant. First variant integrates all components of the viewing
geometry and sensor as well as the Earth and cartographic
projection. In this variant from 5 to 12 well identified and
distributed control points was measured. Accuracy of the
corrections has been checked on well defined check points.
Results are shown in (Tab. 1).
Number of control
Number of check points
RMSE on the check points in meters
points for test area
for test area
Warsaw test area
Cracow test area
Warsaw/Cracow
Warsaw/Cracow
X
Y
X
Y
12
16/12
0.44
0.46
0.44
0.46
! 10
18/14
0.45
0.47
0.46
0.44
8
20/16
0.46
0.48
0.45
0.46
6
22/18
0.53
0.48
0.57
0.53
5
23/19
0.73
0.82
0.84
0.92
Table 1. Geometric correction results of the Resurs DK-1 images in first variant.
Model of geometric correction of the Resurs DK-1 scenes based
on first variant is sufficient for obtained results below half pixel
of the source image. The accuracy on 8 independed check
points was RMSE X = 0.45m and RMSE Y = 0.46m. Using the
orbital parameters of the Resurs DK-1 satellite caused the
limitation of measure of control points required for geometric
correction. The results obtained in this variant show, what is
influence of real orbital satellite parameters on the
mathematical model of geometric correction of the Resurs DK-
1 images. Obtainment of the results of geometrical correction
below one pixel of the source image is possible thanks to use
just 5 control points. Enlargement of the number of measured
control points to 8 let’s get in this variant the optimum result.
The second variant of the geometric correction uses a number
of ground control points. Unknowns of Rational Polynomial
Coefficients (RPC) were calculated on the basis of ground
coordinates of control points measured in the field and on the
image. The accuracy of RPC parameters depends on the
number of control points, accuracy of their identification in the
field and on the image, and their distribution on the scene. RPC
data have been estimated independently from different number
of control points. The results of geometric correction are
shown in (Tab. 2).
Number of
control points
Number of check
points for test area
Warsaw/Cracow
Number of
polynomial
coefficient
RMSE on the check points in meters
Warsaw test area
Cracow test area
X
Y
X
Y
24
4/0
12
0.32
0.36
0.37
0.36
22
6/2
11
0.34
0.32
0.36
0.38
20
8/4
10
0.40
0.38
0.39
0.38
18
10/6
9
0.45
0.45
0.41
0.44
16
12/8
8
0.52
0.46
0.48
0.45
14
14/10
7
0.58
0.62
0.49
0.55
12
16/12
6
0.76
0.82
0.69
0.71
10
18/14
5
0.82
0.90
0.88
0.97
8
20/16
4
1.04
1.18
1.22
1.34
Table 2. Geometric correction results of the Resurs DK-1 images in second variant.
Delimitation of the optimum degree of the polynomial for
estimation of the relationship between image matrix and
geodesic coordinate system was the principle in analyzed
variants. Obtainment of the result of the geometrical correction
on the level of standard deviation of measure and identification
of check points is possible during delimitation of 9 polynomial
coefficients on the basis of measure 18 control points.
Enlargement the number of control points needed for determine
of RPC give the result of geometric correction RMSE X =
0.32m and RMSE Y = 0.36m. However in this case the
necessity of measurement as much as 4 control points disqualify
this variant of geometrical correction. On the basis of results
analysis in the individual variants of geometric correction was
affirmed that obtainment of the accuracy below one pixel of the
source image guarantees the measurement only 10 control
points evenly distributed on the scene. Taking into account the
economical aspect of geometrical correction of Resurs DK-1
scenes in particular possibility to minimum of limitation of
expensive field measurements it should to use the first variant
exclusively. However his final choice is determined by the
access to original orbital sensor data. Additionally analyzing a’
priority and a’ posterior errors on the control points it was