/5 
/3 
Thematic information extracted from 
LANDSAT images - forest  physiognomy, 
anthropism and cloud cover - is recorded 
within each map sheet for each basic 
cell. The date of the image used is 
assigned as an attribute for... both 
anthropism and cloud cover, because they 
are time-dependent parameters. The 
analysis of the LANDSAT images in 
conjunction with the vegetation map of 
Brazil allowed the separation of four 
classes of forest physiognomy: primary 
forest, secondary forest, non-forest and 
water (Meira Filho, 1991). 
As described before, the conventional 
photointerpretation approach was 
considered unfit regarding an accurate 
insertion in the database of the 
information extracted from LANDSAT 
images. Both methods #1 and #2, which 
descriptions follows, are based on the 
same alternative photointerpretation 
procedure. The method consists in 
avoiding that first conventional step of 
transferring . map corners, rivers and 
roads from map to the overlay prior to 
the extraction of image content. The 
interpreter acts directly over the image 
without any information from cartographic 
base. Besides the basic information 
(anthropism, forest physiognomy and cloud 
cover), rivers and roads are also 
extracted from the image. These features 
provide the control points that are used 
by methods 4l or #2. The idea is to 
produce an overlay that is a perfect copy 
of the image in terms, of geometry, 
allowing then the modelling of image 
content through an affine transformation 
during digitizing. 
2.2 Method #1 
Method #1 can be considered as a "direct" 
input of image content (overlay) into the 
1:250,000 map sheet projects. The process 
starts with an approximate subdivision of 
the whole overlay by map sheets, as shown 
in figure 3. 
Figure 3 - Overlay divided by map sheets 
This approximate partition is carried out 
by a different group than. that. of 
photointerpreters, and has the objective 
of selecting roads intersections and 
rivers confluences for control points. 
This method requires a minimum set of 
four control points for. each individual 
Subdivision. This number of control 
points represents the minimum requirement 
to produce a least square estimation of 
the parameters of an affine 
transformation. The existence of one set 
Of control points for each subdivision 
explains why the method is called 
"direct". Each subdivision of the overlay 
is mapped directly into the corresponding 
703 
  
  
  
  
  
  
map sheet project, as illustrated in 
figure 4. 
E 
] * v E. 
M e A 
i. * ! 
+ + A +[ 
  
  
  
  
  
Figure 4 - Direct mapping with different 
sets of control points 
As this method deals with approximate 
subdivisions, overlay data are digitized 
exceeding the boundaries of the cells 
which they are mapped into. The last step 
is then an automatic clipping procedure 
by the cells over digitized data in order 
to determine all intersections between 
lines and cells boudaries (nodes) and to 
delete exceeding arcs. Figure 5 shows 
this situation. 
  
Figure 5 - Clipping procedure 
2.3 Method #2 
Method #2 is a variation of the first one 
and can be considered as an "indirect" 
method. Instead of using different sets 
of control points, only one set is 
required for the whole image overlay. 
Again, due to least square estimation of 
parameters, a minimum set of four control 
points is required. The entire overlay is 
digitized and stored in a temporary file, 
which is further used for the tranference 
of data into the map sheet projects. 
Therefore, the method is called 
indirect”... Figure... 6... 1ılustrates. ‚the 
Situation ‚of. control points. in this 
method. 
Figure 6 - Indirect method and control 
points 
Overlay data are stored in the temporary 
file in terms of table coordinates (e.g., 
mm) and are mapped into each map sheet 
project by means of the geodetic 
coordinates of the control points through 
an affine transformation. The last step