Input Processing Output 
Satellite . 
Image Data = Preprocessing 
I 
Ground Geo metrical 
» Mosaicking 
Control Points 
  
and Rectification 
1 
  
  
  
Radiometrical 
Mosaicking 
  
  
I 
Combination 
of Multisensor 
Data Sets 
I 
  
  
Postprocessing 
  
I 
  
  
  
  
  
  
  
  
  
  
  
Topographic Cartographic 
Informations “| Processing 
T 
Rasterplotter _ | Originals for 
Printing Offset Printing 
  
  
  
  
  
  
Fig. 1: The production line for Satellite Image Maps 
The combination of several scenes and especially mer- 
ging of multisensor data requires very precise geo- 
metrical mosaicking and rectification. For this purpose 
the well-known bi-variable polynomials based on the 
coordinates of some ground control points (GCP's) are 
appropriate. However, the best results can be achieved 
if all transformation parameters for a block of scenes 
are determinded simultaneously in a least squares 
adjustment. In order to reduce the number of GCP's 
and to improve the quality of the result it is advisable 
to introduce tie points (also referred to as transfer 
points) into the adjustment. This approch, already 
proposed in 1984 (Milkus, 1984), yields excellent 
results, and became popular in the meantime (Galtier 
et al., 1992; Rivereau et al., 1992). 
If mountainous regions are concerned the influence of 
the topographical relief must be considered (Albertz et 
al., 1990). For this purpose the rectification based on bi- 
variable polynomials has been extended by approxi- 
mation formulas correcting for displacements caused 
by heigth differences. The approach developed is based 
on the fact, that relief displacements only occur in the 
direction of the scan lines. A height difference Ah (re- 
lative to a mean terrain height) causes a displacement 
As, resulting in an incorrect position of the particular 
pixel (Fig. 2). With regard to rectification and mosa- 
icking the relief displacements are of influence in two 
Ways: 
— The coordinates of the ground control points mea- 
sured in the image data are incorrect. Therefore the 
computation of the transformation parameters is 
based on inaccurate data. 
— The positions in the input image data computed 
during the resampling process are influenced by the 
actual terrain height. 
Both effects accumulate so that the elemination of 
height-caused displacements becomes an important 
problem. If the displacements can not be tolerated and 
must be corrected in a particular mapping project, the 
availability of a Digital Elevation Model (DTM) is 
essential. 
sensor 
mean terrain level 
  
level of reference system 
  
AS: : scan line 
  
Fig. 2: Displacements As caused by terrain height 
differences Ah 
For single scenes, the measured ground control point 
coordinates can be corrected by use of the appropriate 
terrain height. However, this is impossible in the case 
of tie points, because of the fact that usually the 
reference coordinates of tie points are unknown before 
the computation of transformation parameters. There- 
fore it is neccessary to develop this procedure into an 
iterative procedure. This is achieved in the following 
way. The first adjustment is carried out with un- 
corrected image coordinates of the tie points. Thus 
approximate values of the tie point coordinates will be 
made available. Through these coordinates a first 
height correction can be determined. Further itera- 
tions will then improve the previously calculated tie 
point coordinates until final results are achieved. 
From the DTM the terrain height for each pixel can be 
provided for the rectification. For this procedure the 
following method, operating in three steps, has been 
developed and successfully applied: 
— At first the equations for indirect rectification lead 
to a position in the input data for each pixel of the 
rectified image, without taking care of terrain height 
influences. 
— After that the displacement correction As has to be 
added so that the correct position is determined. 
— Finally the gray value for the correct image position 
is interpolated and transferred to the rectified image 
data (resampling procedure). 
According to this approach the problem of rectification 
of satellite image data considering the terrain relief can 
be solved with high accuracy. Nevertheless, it should 
be emphasized that accurate rectification can only be 
achieved if precise and extensive height information 
in the form of a DTM is available. 
After the geometrical rectification also radiometrical 
mosaicking is essential, because the radiometry of the 
image data differs significantly between adjacent 
scenes due to a variety of effects (sun elevation, 
310 
  
  
  
ACTA Em ol SN 
nd x 0. <# ner …no + 
m4 8 (S Ll PS t9) er UN 0, hà bete 
fe» 
Pde M 
QT ( S orm