applied for rectification has to be extended by using an 
actual DTM (ALBERTZ et al. 1990). 
It must be considered that scanners have another 
geometrical model than aerial images. In this cases the 
relief displacements mainly occur in the direction of 
the scan lines. 
The influence of the relief to rectification and mosaic- 
king can be described as follows: 
The coordinates of the measured GCP's in the 
image data are not correct. This means the calcu- 
lation of the transformation parameter is based on 
incorrect input data. 
The positions of the input data for the resampling 
process are influenced by the actual terrain hights. 
Both effects accumulate so that depending on the 
scanner and the relief differences the displacements 
cannot be without consideration. In this case the 
measured GCP's must be corrected by actual terrain 
heights. This is not possible for tie points, because the 
absolute reference coordinates of tie points is not 
known. Therefore the final coordinates of tie points 
are calculated in an iterative process. 
In the final rectification for each pixel a correction As 
can be calculated by using the DTM. This correction 
will be considered for the correct image position of the 
resampled gray value. It must be emphasized that a 
precise rectification needs a DTM with high accuracy. 
24 DTM-Generation from Satellite data 
It is obvious that a DTM is one necessary layer in a GIS. 
However, particulary in developing countries the 
availibility of DTM's cannot be assumed. In this case 
the generation of a DTM with stereo-images takes 
place. Especially stereo-SPOT data play an important 
role in this context. With the capabilities of SPOT- 
satellite DTM's with height accuracies better than 10 m 
can be achieved. 
The photogrammetric evaluation of stereo image data 
is usually divided into three steps: interior, relative 
and absolute orientation. These procedure has to be 
adapted to the geometrical properties of SPOT data. 
The algorithm used in this software package is descri- 
bed in ALBERTZ et al. 1990. 
However, the basic tasks of automatic DTM production 
is the determination of homologue points using 
special computer algorithms and hardware. For image 
matching the combination of two methods is in use: 
the normalized cross correlation for the calculation of 
approximation values and the least-squares matching 
for final results with subpixel accuracy. Experimental 
results with SPOT stereo-pairs have shown that accu- 
racies better than 10 m can be achieved. The elevation 
of DTM's from SPOT satellite have provided height 
data with sufficient accuracy - reliable for the rectifi- 
cation of satellite data and the integration into GIS. 
Furthermore KFA-1000 data offer also the possibility of 
evaluation of stereo-data. Due to the slight better reso- 
lution and a smaller base-height ratio at least similiar 
results can be expected. 
690 
2.5 Radiometrical Mosaicking 
The result of the geometrical processing is a data set of 
several rectified images with high accuracy but which 
are single files. These scenes can differ significantly in 
radiometry due to different acquisition dates, atmo- 
spheric effects, etc. Hence it is necessary to compose 
one image within a common gray scale system. In 
order to acquire a homogeneous mosaic the infor- 
mation content in overlapping regions of adjacent 
scenes is used again (KAHLER 1989). 
The procedure starts with the definition of the over- 
lapping regions which have to be used for the 
algorithm. After this step median integral histograms 
are calculated for each double information. The result 
is used to calculate correction tables for each scene in 
an iterative process. Afterwards a data set is available, 
which presents all scenes in an homogeneous gray 
scale system. The last step is the elimination of the 
double information in order to achieve just a single 
file. For this purpose special transition algorithms can 
be defined. Practical experiences have shown that a 
square transition related to a defined separation line 
between two scenes yields the best results. However, 
sometimes remaining effects can occur if regions with 
very different gray values be adjacent. In this cases 
some interactive corrections have to be applied. The 
method has been used very often in the last years and 
yields excellent results. 
2.6 Combination of multisensor data sets 
In order to make available the full information 
content of satellite image data it is often very useful to 
merge the advantages of different sensor image data. 
In particular the combination of panchromatic SPOT- 
data with multispectral Thematic Mapper (TM) data 
provides high geometrical resolution merged with a 
lot of possibilities in spectral band combination. This 
offers wide application possibilities for various map- 
and GIS-users. 
MULTISPECTRAL 
DATA 
DATA 
CHANGING 
new 
intensit 
eem 
COMBINED 
DATA 
    
   
  
[res] 
Fig. 3: Principle of combination multisensor data 
The best results for the combination of multisensor 
image data can be achieved by using the IHS color 
transformation (TAUCH & KÄHLER 1988). The principle 
of. this method is explained in figure 3. After the 
transformation of the original multispectral data set 
(RGB) the intensity, hue and saturation components 
(IHS) are provided. Now the substitution of the inten- 
sity from the original data with the data of high