dium den-
cted in
oth clas-
ighbour-
idential
ation also
ry fre-
es the
Ward coun-
: the town
7 occur
ates a
ng-yards
four clas-
ver, the
| but, in
| significant
icis). Final-
rsides (Pa-
of land
ining the
similar to
f the Uni-
g" around
be completed
is case,
ork of the
. the image
= 89/ -
showing a building site, or to the linear featured scheme they form when showing
a highway under construction. So the utility of pattern recognition appears evi-
dently which was formerly perceived in photo-interpretation (Fig. 3).
In the Charleroi area, the irregular pattern of the urban districts
makes the interpretation difficult. The classification level of discrimination
applied to the recording of March 1973 enables to detect visually some spatial
associations of classes. In several instances these asso-
ciations correspond to. a specific land use. So the density gradients of the
urban space assigned to dwelling appear clearly. The main urban core corres-
ponds to a presence of classes 3 and 7 of which cumulative surface is greater
than 85 % of the total. Classes 7,5, 8 and 9 are typical of minor cores (total
greater than 85 7%) In the blocks of traditional dwelling in-
cluding yards in their center, these four classes total up 70 to 80 %. The
lower density settlement has a more complex spectral signature, where classes
7, 8 and 9 are the more frequent. The detached houses in wooded plots areas
always include an approximative rate of 60 Z of classes 8 plus 6.
Other spatial sets correspond to objects having a similar reflec-
tance but various shapes and functions. This is the case for all areas compri-
sing workshops or warehouses covered with bright roofs, building sites and
wide transportation infrastructures. After assistance the 4 corresponding classes
(fig. 4), take up at least 80 Z of the surface. Some objects have
a variable signature, and can be identified by their shape and environment.
For example, the coal refuses include families of various reflectances according
to the sort of material and the rate of vegetal cover, but most of the time can
be marked out from their built up surroundings. Heavy industry is the type of
land use the most difficult to characterize. It occupies large parts of the
urban area. Some of them have the same signature than thickly settled districts,
others involve unusual proportions of classes 1 and 3, other are wholly hidden
by the smoke outlets. Classes 5, 7 and 11 seem to be the best indicators of
the heavy industry presence.
b. Specific classification problems for urban and industrial areas
The recordings of March 22, 1973, for Brussels and Charleroi partly
undergo a disturbance of their reflectances in the four channels of LANDSAT.
It obviously is noticeable in the multispectral analysis : the origin of such
a phenomenon lies in the atmospheric pollution of definite areas in these urban
agglomerations.