Andrade, Nilo Sergio de Olive 
  
results of the analyses can also be seen in Figure 7. It must be noticed that the effectiveness of the supervised 
classification is dependent of the number of pixels included in the training samples. These pixels will be used to 
estimate the mean vector (m) and the covariance matrix (S) of the whole spectral class. Considering x the vector 
correspondent to a pixel in the N classes involved, the mean vector of the pixels belonging to a class is given by: 
1 k 
m=E@= 2 1) 
Where K is the number of pixels in the class and E(x) is the expectancy of x. 
and the covariance matrix is given by: 
S. = firm} emir) (2) 
  
sas 
-— 
a oO 
  
Changed Area (Km2) 
o 
Red White Blue Yellow Cyan Black Brown 
Color Used 
—e— ground truth —#— K-means —14-— Maximum likelihood 
  
| 
  
  
  
Figure 7. Results obtained with the visual analysis of the TM images, as well 
as with the supervised and unsupervised classifications. 
5. CONCLUSION 
This study reported data on landscape changes between 1995 and 1997 for a region that has undergone important 
alterations during the last three decades. Results were based on a relatively simplified approach that nevertheless 
showed how these changes are causing the continuous forest depletion in this region. 
This continued process of forest removal is possibly happening in other areas of the Amazon and the use of remote 
sensing and spatial techniques can help the identification of such a process and contribute to a better understanding of 
the processes occurring in the Amazonian region. The supervised classification showed to be a good tool for this 
purpose. 
6. REFERENCES 
Alves D.S., da Costa W.M., Escada M.I.S., Lopes E.S.S., de Souza R.C.M., Ortiz J.D, 1998a. Uma anälise das regiöes 
que apresentaram as maiores taxas de desflorestamento na Amazönia Legal nos periodos 1991-1992 e 1992-1994. Säo 
José dos Campos: Instituto de Pesquisas Espaciais, 1998 (Technical report INPE-AMZ-R03/97). 
Alves D.S., Pereira J.L.G., Sousa C.L., Soares J.V., Yamaguchi F., 1998b. Classification of the Deforested Area in 
Central Rondónia using TM imagery. [CD-ROM]. In: Simpósio Brasileiro de Sensoriamento Remoto, 9., Santos, 1998. 
Anais. Säo Paulo: Imagem Multimidia, 1998. Seçäo de Posters. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 
69