2-2 Density Transformation Method 
The following four typos of density 
transformation functions were examined to 
SPOT data of Japan and Peru. 
(1) Linear contrast stretch function 
(2) Multi-zonal linear functions 
(3) logarithmic function 
(4) Histogram flattening function 
Data distribution of Peru scene was rather 
normal but mostly below level 70. Simple 
linear function to stretch gray level 
range to full 256 levels gave very clear 
image. On the other hand. Japan data was 
irregular as bi modal distribution. 
Therefore, single linear function could 
not correct whole range of data. 
Different slope linear functions are 
necessary for each lumps of data. Then, 
dual linear "contrast stretch function was 
adopted and gave good result. 
it is supposed to give good result that 
original gray level is divided into a few 
zones according to "data. histogram : and 
different linear functions are adopted to 
each zone. 
3. PLOTTING ACCURACY 
SPOT data is stated to be applicable for 
1:50,000 scale topographic mapping. |n 
accordance with the specification adopted 
in international cooperation project of 
Japan, planimetric error should not exceed 
35m and altimetric error should not exceed 
7m in class À mapping of. the scale. of 
1:50,000. Since the pixel size of SPOT 
data is 10m, 35m planimetric accuracy is 
seemed to be accomplished in most cases. 
However, 7m altimetric accuracy might be 
rather hard to be achieved. Therefore, 
height measurement accuracy of SPOT stereo 
pair was examined. 
3.1 Examination 
Three SPOT images of Mt. Fuji area were 
used as test images. The specifications 
of the images are as follows. 
Column-Low 929-279 
Observed date Mar. 7, 1986 (Center) 
Mar. 8, 1986 (Left) 
Mar.17. 1986 (Right) 
Off nadir angle 4.0 East (Conter) 
24.1 West (Loft) 
15.7 East (Right) 
Mode Panchromatic 
Processed level 1A 
SPOT image orientation and measurement was 
carried out by using PLANICOMP C=100 
analytical plotter and the SPOT plotting 
software "BINGO". 32 GCPs were selected 
from 1:25,000 scale topographic maps for 
orientation of images. 
9km of 
every 
images 
the two test sites of 3km by 
and mountainous area, DEM of 
were measured from oriented 
compared to DEM measured from aerial 
Furthermore, contour line of 
was drawn at the scale of 
At 
flat 
100m 
and 
photographs. 
40m interval 
337 
1:50.000 and compared to 1: 50.000 
topographic map. 
3.2 Height measurement accuracy 
Orientation accuracy at the 32 OOPs was 
such that planimetric accuracy is 5m both 
in Center-Right pair and Left-Right pair, 
and altimetric accuracy were bm and 4m in 
Center-Right and Left-Right respectively. 
Orientation accuracy is rather constantly 
good as decreasing number of control 
points. Required orientation accuracy is 
seemed to be achieved if there exist some 
ten control points. 
Table 3-1 shows the result of DEM 
comparison. Root mean square of elevation 
errors were within 7m tolerance except the 
of mountainous area of  Center-Right 
case 
pair, but quite close to 7m tolerance in 
other cases. From this result. it is 
hardly to say that 1:50,000 scale maps can 
constantly be produced out of SPOT images. 
1:100,000 scale maps with 40m interval 
contour lines are supposed to suit to SPOT 
data mapping. 
  
  
  
  
  
  
  
  
  
  
  
  
Table 3-1 Residuals at DEM points 
area pair | B/H | RMS max. 
plain C-R | 0.52 | 5. 5m | -24. 7m 
plain L-R 0.72! 5.8n |. 22.4n 
mountain | C-R | 0.52 | 9.6m | 62. 7m 
mountain | L-R | 0.72 | 6.9n | 31. Im 
Figure 3-1 of the next page shows the 
contour lines drawn from the SPOT stereo 
pair, while figure 3-2 shows the |:50.000 
scale topographic map of the corresponding 
area. Through the visual inspection of 
two figures. it was. convinced that 
principal terrain features as main ridges 
and valleys were correctly plotted as 
sufficient for 1:100,000 scale topographic 
maps. 
4. CONTROL POINT EXTRACTION METHOD 
In aerial photogrammetric mapping process, 
anti-photo signals were set at 
triangulation points before photographing. 
On the other hand, in the case of SPOT 
mapping, pt. Ise quite difficult to set 
anti-photo signals in advance of imaging. 
Therefore, it is necessary to develop 
practical method to extract ground control 
points out of given SPOT images. 
technique 
much higher 
required in 
image 
GCP extraction 
in remote sensing. 
accuracy of GCP coordinates 
mapping than in remote 
registration. 
is quite common 
However, 
is 
sensing 
points for mapping must 
geodetic coordinates and 
identified on an image. 
available, geodetic 
coordinates can easily be obtained any 
place in the world. The difficulty ..is 
that there can not always exist clearly 
identifiable and pointable ground object 
around where we need a GCP. Here, two 
control 
correct 
should be clearly 
Since GPS got 
Ground 
have