Noernberg, Mauricio Almeida 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
HETEROGENEOUS 
HH (Normal) VV (Normal) 
T =. T PTE rer TITIES RE pre pr ro 
e 9 s 
0.020 se = $ 
$e 7 0.06 + 
> i 
* 32 4 
0.015} - 
0.04 À 
0.010} i - 
; 
| > % i 0.02 2 
0.005 |- = 
= + 
0.000 i 1 tas 0.00 
50 100 150 5o 
HV (Normal) VH (Normal) 
Errrerer TT'TTTTTT bi] " TTTTTTT TTTTTTTTTT TTTTTTTTTT [UTE 0 à Seis Sei Ca AIDE I TITTY TTTTTTTTT "T = TTTTTTT TerTTTTTTT "TTTTTTT T 
$ 
à 
0.56 - 4 0.06 - - 
e e 
o 
+ e 
E 
0.D4}H = = 
o 
+ 
+ 
$ 
0.02 - $ - - 
$ + 
Ll 
es 5 
e 
o 
a.no LL bL LEA Leda siad 
10 20 30 40 50 50 
  
Figure 6. Statistical distributions associated with the Heterogeneous class, at each polarization. 
5 CONCLUSION 
The use of conventional classifiers such as the maximum likelihood (Maxver), which assumes a normal 
distribution for the data, may not be adequate for SAR data in general. This suggests the importance to develop 
new digital classifiers that take into account distinct statistical distribution for the data, thus allowing for a larger 
range of radar data applications. 
ACKNOWLEDGMENT 
The authors thanks the Brazilian National Institute for Space Research, the Canadian Centre for Remote 
Sensing, the Canadian International Development Agency, and the European Space Agency for supporting this 
work. Thanks are also due to Centrais Elétricas do Norte for the support during the ground data collection in the 
Tucurui reservoir. 
REFERENCES 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1017