Grown Control Points - Georeferentiation
Image 229 - 079
GPS (WGS 84) Coordinates to Gauss - Kiiger
Rectification - Errors
Data
G. C. P.
Grown C.
Points
Gauss - Krüger Coordinates
Image Pixel Coordinates
Residual Error
R.M.S.
(Root Mean Error)
Column
Row
X(nord)
Y (est)
X
Y
X
Y
9
7.056.623,949
4.440.205,522
857,328
5.451,453
0.363
0.127
0.385
10
7.047.704,583
4.523.857,638
3.609,141
5.600,547
0.089
-0.029
0.039
12
7.001.107,811
4.437.845,464
1.071,516
3.657.547
-0.936
0.055
0.938
8
6.963.367,212
4.437.000,970
1.241,734
2.439,920
0.239
-0.033
0.241
6
6.897.860,137
4.401.211,035
427,266
149,859
0.286
-0.177
0.336
4
6.908.868,274
4.538.027,734
4.793,281
1.215,844
-0.267
0.356
0.445
13
6.947.852,357
4.555.307,808
5.148,672
2.557,453
0.028
-0.206
0.208
11
6.978.417,342
4.578.022,819
5.723,031
3.656,969
0.314
0.052
0.319
7
7.016.618,837
4.597.134,886
6.141,391
4.983,703
-0.116
-0.145
0.185
Errors in Pixels
Root Mean Square in X
0,16510
Root Mean Square in Y
0,38572
Total Root Mean Error
0.41956
TABLE 3
If the rectification process, with the calculated transformation
matrix gives a R.M.S. lower than !4 pixel, the program enter
to the rectification dialog box. To calculate these errors, the
system, with the trasnformation matrix calculated, obtains the
rectificated coordinates of all the GCPs used, then carries out
an inverse transformation and compare the pixel coordinates
obtained (x r ,y r ) with the original pixel coordinates of any point
(Xi.y,),
With the expression
>l( x r -*>)’+O', -y,Y
Total R.M.S. = ^R a 2 +R/
the R.M.S. for any point is obtained , also :
(See Tables 2 and 3)
R¡ =VxR, 2 + YR, 2
In the Rectification Dialog Box, it is possible to
choose the resample method and the pixel size of
the rectificated image to be obtained.
where
Ri = Root Mean Square of the Grown Control Point “i"
XR| = residual in X of the GCP T
YR| = residual in Y of the GCP “i"
To calculate the R.M.S. in X and Y, and the Total RMS, the
following expressions are used:
The resample method is the program procedure
to establish the digital levels (DL) to asign to the
resultant image pixels. The software offers 3
alternative methods:
1. - Nearest Neighbour
2. - Bilineal Interpolation
3. - Cubic Convolution
We selected Nearest Neighbour that the software
offers as default. This method places in any grid
location of the corrected image, the DL of the