wo 
^ia 
  
MAPPING FROM MULTISENSOR STEREO IMAGES WITH ANALYTICAL PLOTTER 
R.J.Ackermann, G.R. Venkatesh 
ITC, Department of Geoinformatics, Enschede, the Netherlands 
ABSTRACT: 
The paper describes an approach for the use of multisensor stereopairs vith an analytical plotter and 
analyses planimetric and height accuracy. 
The potential of mapping from satellite imagery has been investigated during the last decade, especially 
since the launch of SPOT, with its 
may offer the best solution today for 
high spatial resolution and stereo capability. 
data compilation, by either a digital 
Although stereo SPOT 
or analytical approach, 
multisensor stereo images may become interesting alternatives. 
A SPOT image, preferably an oblique view, can be associated vith a stereomate (Landsat-5 TM or IRS India) 
which requires preprocessing in order to generate a pseudo SPOT stereopair. 
The low B/Z ratio will affect mainly the height accuracy, while the planimetric accuracy is comparable to 
results obtained by normal stereo 
order of#60 m, which are inherent to the TM 
resulting from the scanning mirror system. 
INTRODUCTION 
The mapping potential of satellite imagery has 
been extensively investigated during the last 
decade. With the launch of SPOT in 1986, stereo 
capability has been added to satellite images, 
which gives an improved visualization, facilitates 
the interpretability of topographic features and 
allows for stereo measurements. Many users try to 
exploit the complementary aspects of various 
satellite data. Merging, for instance, a 
TM-Landsat with a SPOT-Pan image will add the high 
spectral resolution of the first to the high 
spatial resolution of the second. This approach 
has been widely used during recent years (Ehlers 
1988). By generating multisensor stereo data 
sets,we add the possibility to create a DEM, which 
can be used for many applications (Welch, Jordan 
1990). The basic idea here is to generate a 
stereopair, comprising a SPOT oblique view and 
another high resolution satellite image (generally 
vertical) and to make use of special SPOT software 
which is now available on most analytical 
plotters. 
Some preprocessing is required since both images 
have quite different geometric characteristics. 
This approach could allow more flexibility in the 
use of satellite data, creating unconventional 
stereopairs from any set of overlapping images. 
Since many users have already, at their disposal, 
single images, stereo capability could be obtained 
by adding one image (preferably SPOT) for a 
selected area. 
There 
approach 
as serious loss of 
is an economic advantage which makes this 
attractive but this should not outweigh 
accuracy as compared with a 
stereo SPOT pair. Although the digital method may 
be more appropriate in handling such 
unconventional stereopairs, the analytical method 
offers the following advantages: 
- high resolution stereo-observation system 
- high precision measurements 
- available software for stereo SPOT images 
303 
SPOT. The heights of SPOT/TM models show local deformations in the 
images and are 
probably due to some geometric errors 
describes a 
well known 
The present paper 
makes use of 
procedure which 
image processing 
techniques to generate a TM/SPOT stereopair. The 
results obtained in planimetry are very close to 
earlier evaluations of stereo SPOT, but the 
results in height reveal unexpectedly large 
deformations, due most probably to some non 
corrected geometrical errors in the Landsat-5 TM 
image. 
DATA SETS AND STUDY AREAS 
The first tests were carried out with available 
data at ITC. The selected area is located in the 
south of France (north of Marseille) representing 
a changing relief from flat to hilly - and 
mountainous terrain. Good up to date topographics 
maps at 1:25 000 and 1:50 000 scales are available 
for the whole area and they contain plenty of well 
defined features to be used as ground control 
points or check points. The two SPOT images, both 
level 1A, are of excellent quality with no cloud 
cover; they were recorded in October 1989 with a 
time interval of 14 days (Table 1). 
The Landsat-5 TM data available in system 
corrected CCT format were acquired from EURIMAGE; 
they were corrected for radiometric and geometric 
distortions. The TM and SPOT images overlap by 
about 95%, but due to cloud cover in the eastern 
part of the TM images, the useful model size is 
limited to 60 by 40 km. Other data sets were made 
available for the experiments: 
- Hannover (Germany): SPOT-XS and TM having a very 
low B/Z ratio 
- New-Delhi (India) : SPOT-XS, level 1B and IRS 
image (Indian Remote Sensing Satellite), 
processed to a level eqivalent to SPOT 1B. 
In all cases good stereoscopic models were 
obtained but the fact that we use images for which 
different geometric corrections have been applied 
will certainly affect the height accuracy.