Iniernational Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
obtain control points used in geometric correction and 
registration of those TM images. 
2.1.3 Geographic Information System (GIS) 
It was used the GIS SPRING 4.0, a coupled geographic 
information/image processing software developed at INPE 
(Camara et alli, 1996). 
2.2 Methods: 
2.2.1 Multispectral bands choice 
It was tested all TM's multispectral bands in terms of response 
to detect desforested areas and the best combination was 
achieved by using bands 4, 5 and 7, all of them in the infrared 
regions, due to the higher response from vegetation cover and 
bare soils. 
2.2.2 Geometric correction and enhancement contrast 
It was used 15 ground controls points extracted from DSG's 
topographic map, both scenes were converted to the UTM 
coordinate system, using a first-degree polynomial rectification 
algorithm, this procedure yielded a registration accuracy equal 
to 0.9 pixel. The 1987 scene was used as reference to coregister 
2002 scene that assured a good image-to-image registration. In 
order to preserve the radiometric integrity, it was used a nearest- 
neighbor interpolation method. 
After the geometric correction, both scenes were submitted an 
enhancement linear contrast to emphasize the best separation 
between vegetation cover and bared soil. 
The figure 2 shows the TM whole scene of 1987, after the 
geometric correction and liner contrast operations. The red 
rectangle points out the study area. 
  
Figure 2 — Color composition of the 1987 TM's multispectral 
bands (band 7-red; band 5-green; band-4-blue) after geometric 
correction and linear enhanced contrast operations. 
2.2.3 Digital input data of deforestation polygons 
By inputing digital algorithms data of several deforestation 
polygons in vector format, located in both TM images, such 
polygons are related with petroleum prospection activities, e.g., 
seismic survey, prospection gas and oil wells. 
All inputed polygons data were associated with thematics 
classes in GIS as following: 
a) anthropic 87 and 02: referent anthropic occupation in 1987 
and 2002 scenes; 
b) roads 87 and 02: referent small roads/paths in 1987 and 
2002 scenes; 
c) seismic 87 and 02: referent small glades opened in the forest 
to seismic survey in 1987 and 2002 scenes; 
d) general glades 87 and 02: others glades opened in the forest 
for petroleum activities like wells in 1987 and 2002 scenes; 
e) clouds/shadows 87 and 02: referent small areas with clouds 
and their clouds presents in 1987 and 2002 scenes. 
The figures 3 and 4 show the polygons after having their data 
computerized. 
  
Figure 3 — TM 1987: Study area: deforested polygons after 
digital input data tasks 
        
i AN NU RE SRS ow C LI M QM DNE MEL 
gure 4 — TM 1987: Study area: detailed deforested polygons 
show glades of seismic survey, paths, glades for helicopters 
points landing and gas wells. 
    
ies 
Fi 
2.2.4 Spatial Analysis 
Using the information plans with vectorial files of deforested 
polygons for each scene, inside SPRING 4.0, they were carried 
out by special analysis tools: algebra of maps, measures classes 
and Kernel's density estimator. 
Algebra of the Maps j 
The algebra of maps uses special languages for algebraic 
geoprocessing (LEGAL), using focal, local and zone operators. 
370 
  
Interna 
LEGAL 
generat 
occurre 
in the s 
classes 
{ 
// Delin 
Themat 
1987 (" 
Results 
//Recovi 
deforest 
deforest 
// Creati 
results= 
Resoluti 
Scale=6 
/Definit 
results 7 
{ 
"GG87N 
&& ! (d 
"GG87E 
&& det 
"GG02N 
&& ! (d 
"SG87NI 
&& ! (d 
"SG87E( 
deforeste 
"SG02N: 
! (defore: 
hl 
Being: 
GG87N0 
GG87E0: 
GG02N8 
SG87N0: 
SG87E02 
SG02N8” 
Measuriti 
Using tl 
deforeste« 
the area | 
study ai 
transform 
Kernel's 
The algo 
estimative 
plans with 
it generat 
possible li 
  
More 
alli (19 
More 
CAMA