Andrade, Nilo Sergio de Olive 
  
visits. While the information classes for a particular exercise are known, the analyst is usually totally unaware of the 
spectral classes, or sub-classes as they are sometimes called. Unsupervised classification is therefore useful for 
determining the spectral class composition of the data prior to detailed analysis by the methods of supervised 
classification. 
The unsupervised classifier used was the K-Means. K-Means unsupervised classification calculates initial class means 
evenly distributed in the data space and then iteratively clusters the pixels into the nearest class using a minimum 
distance technique. Each iteration recalculates class means and reclassifies pixels with respect to the new means. All 
pixels are classified to the nearest class unless a standard deviation or distance threshold is specified; in which case 
some pixels may be unclassified if they do not meet the selected criteria. This process continues until the number of 
pixels in each class changes by less than the selected pixel change threshold or the maximum number of iterations is 
reached (Tou and Gonzalez, 1974). For this study the number of iterations selected was 100; the number of classes was 
5 and change threshold was 5.00. 
Supervised classification procedures are the essential analytical tools used for the extraction of quantitative information 
from remotely sensed image data. An important assumption in supervised classification usually adopted in remote 
sensing is that each spectral class can be described by a probability distribution in multispectral space: this will be a 
multivariable distribution with as many variables as dimensions of the space. Such a distribution describes the chance of 
finding a pixel belonging to that class at any given location in multispectral space. This is not unreasonable since it 
would be imagined that most pixels in a distinct cluster or spectral class would lie towards the centre and would 
decrease in density for positions away from the class centre, thereby resembling a probability distribution. The 
distribution found to be of most value is the normal or Gaussian distribution. It gives rise to tractable mathematical 
descriptions of the supervised classification process, and is robust in the sense that classification accuracy is not overly 
sensitive to violations of the assumptions that the classes are normal. 
The supervised classifier used was the Maximum Likelihood. Maximum likelihood classification assumes that the 
statistics for each class in each band are normally distributed and calculates the probability that a given pixel belongs to 
a specific class. Unless a probability threshold is selected all pixels are classified. Each pixel is assigned to the class that 
has the highest probability (i.e., the "maximum likelihood"). For this study no probability threshold was selected. 
4. RESULTS AND DISCUSSION 
Regions of Interest were created for each situation (changes in land use and cover) to be analyzed in the images and the 
colors chosen for each one are presented in the Table 1. 
  
  
  
  
  
  
  
  
  
1995 SITUATION 1997 SITUATION | COLOR 
Mature and secondary forest Bare soil Red 
Mature and secondary forest Pasture White 
Bare soil Pasture Blue 
Pasture Bare soil Yellow 
Regrowth Pasture Cyan 
Burned areas Bare soil Black 
Burned areas Pasture Brown 
  
  
  
  
Table 1 — Table of colors used for the visual analysis of the Landsat Images. 
The Figures 1 and 2 depicts the 1995 and 1997 TM images respectively, after the process of radiometric enhancement. 
These images were visually analyzed and a color mask representing the changes in land use/cover obtained. This mask 
was superimposed to the image acquired in 1997 as is shown in Figure 3. Parts of the region's landscape can be 
appraised on Figure 4. | 
After the analysis and interpretation of the images it was performed the calculation of the area corresponding to each 
and every one of the changes identified. The total area of each change was calculated multiplying the area of each pixel 
(0.9 Km?) by the number of pixels present in the Region of Interest (ROI). Percent of change is the modified area 
relative to the total study area. The results can be seen in Table 2. 
  
  
  
  
  
  
  
  
1995 SITUATION 1997 SITUATION | CHANGED AREA | PERCENT OF CHANGE 
Mature and secondary forest Bare soil 29.79 km“ 2.71% 
Mature and secondary forest Pasture 40.00 km“ 3.64% 
Bare soil Pasture 33.47 km“ 3.05% 
Pasture Bare soil 4.58 km” 0.42% 
Regrowth Pasture 1.26 km” 0.12% 
Burned areas Bare soil 1.73 km” 0.16% 
Burned areas Pasture 0.16 km? 0.02% 
  
  
  
  
  
  
Table 2 —Evolution of changes in Land use/cover. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 65