Version 2 using the affine transformation
was not better than version 5 as was initially
expected, probably due to the poor definition, and thus
pixel-coordinate accuracy, of the GCPs. However, even
if the accuracy was only similar, the number of needed
GCPs, and the associated costs and efforts, were much
less. Note also that although the check points had much
different heights than the four used GCPs (see Section
5), no height extrapolation errors occurred. Version 3 is
like version 2 but only with the points that were well
identifiable and measurable, and shows the accuracy
potential with good GCPs. The larger RMS and maximum
absolute error in X for all versions is expected
(terrain influence is in CCD-line direction, which is
closer to X). The achieved accuracy is better than for the
Precision product of Space Imaging, while the costs of
the latter are ca. five times higher than Geo, and its
radiometric quality worse than our postprocessed
orthophoto.