epeated 
s. The 
ields) 
Or 
age 
signa- 
e 
as are 
ness 
t 
na 
rs and 
tnt 
. The 
orted 
dow 
w of 
soil 
t. With 
O 
mask 
idensity 
land, 
ised 
cal 
ing 
to 
on, 
O 
p 
esti- 
the 
nd 
and 
wn 
maps. 
map; 
  
on. 
P 
GENRES i cam m 
(the range of var tion is less than .15cm less or more) we can 
estimate : the scale of each image after the regional analysis 
procedure according to the criteria of spatial differentiation 
(the inverse of the above mean width). The results of our eight 
experimental sites are presented in figure 8 below. 
  
  
  
Scale 
A 
1/400,000 B 
1/300,000 - 
1/200,000 À 
- 
1/100,000 2 
= 
© Size of Window 
j| v y V y ) . 
11/ 13 ie. 20 .33 44 
Figure 8: Relation between the estimated scale by the 
mean width of unit and the size of window used in the ICAR process 
Although eight points are not enough for statistical test, 
it can be seen that these points form a straight line whose eauat- 
ion is Scale » ll. Size+105 (R=0.9 ) 
Other experiments have to be done to confirm this which may be of 
great help in answering the questions such as the best size of 
window to produce a map of a given scale ? 
  
Even if the results are positive, other problems arise 
concerning: 
. a more complete description of the landscape, using geometrical 
parameters, that may be necessary with high resolution satellite 
. the place of such a study in the planning of a survey. Is it 
necessary to apply it before the survey, as a first reconnaissance 
of the area, during it, or at the end as a help in the drafting of 
boundaries. 
Irrespective of cost, the development of specialised computer 
for image processing will make these processes more widely 
available because of their use in mapping, in regional knowledge, 
and in stratification for agricultural statistics. 
Bibliographical references: 
1 Rosenfeld A 1970, Connectivity in digital pictures. Journal 
199 
= = = BR III 
uM REBEL EMBED A —