around Aix-en-Provence 
Fig. 3: Stereogram showing the height deformations 
Analysis of the profiles shows a systematic 
pattern of deformation; yet some changes in the 
magnitude or sign of the systematic errors occur 
often every 5 to 6 intervals of 500 m. 
Converting a height error into a parallax error, 
applying, with some approximation, the formula 
by 
used in photogrammetry, we write: 
and AZ-60.0 m Ap » 30.0 m 
corresponds exactly to a one 
: A parallax of 30 m 
— pixel shift in the original TM. Apparently we are 
EY confronted here vith an old problem of pixel shift 
= (1-1.5 pixels) which becomes critical only if we 
= want to use Landsat-5 for height 
- measurements and DTM generation. 
  
  
TM data 
CONCLUSION 
The present approach for the use of multisensor 
images offers an interesting alternative next to 
Stereo SPOT. Accuracy performances obtained with 
stereo TM/SPOT compares well with stereo SPOT 
images heights show an 
irregular 
values up to + 
deformations can be easily located, especially in 
flat terrain. Unless some improvement can be made 
of the Landsat-5 TM 
the 
in planimetry, but the 
of deformations with maximum 
of 
pattern 
60 m.Some of these areas 
in the geometric processing 
T & 
P rend 
—= data, to correct for some system errors, 
E pea height inaccuracy will be the limiting factor for 
eas mapping applications. 
TM 
x There is actually a growing interest for stereo 
images and there will be new satellite images 
available in 1992 (e.g. Landsat-6 ETM) with new 
possibilities for using multisensor stereo images. 
POT 2 
approach for handling such 
believe that 
  
>= 
Although the digital 
sets seems more appropriate,ve 
data 
for the time being and for some years to come, the 
analytical plotter with its SPOT software will be 
used for data acquisition from satellite imagery. 
  
  
  
nce 
ON, 
can 
m, 
uts 
sly 
Fig. 4: Height deformations 
two 
Aix-en-Provence 
along profiles near 
IS- 
can 
307