International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004 
  
channel mean st. deviation 
water (from 12792 samples) 
93.35 
62.47 
37.6 
urban area (from 5548 samples) 
Green (71/2) 125.23 
Red (7/2) 99.76 
NIR (mf2) 88.40 
Green (7/3) 112.2 
Red (7/3) 76.20 
NIR (7/3) 197.66 
- crop 2 (from 3461 samples) 
Green (7/4) 121.82 
111.01 
124.50 
Green //n/5) 76.57 11.90 
Red (7/5) 47.37 12.34 
109.88 28.03 
  
Table 1. Mean and standard deviation values of training areas 
As it can be seen in following figure, similar values (overlap) 
can be found in the green channel for crop 1, crop 2 and urban 
area classes. This is due to the similar characteristics in the 
spectral response (reflectance) of these classes in the 
wavelength range 0.5—0.59 um. Fortunately, they can be better 
separated cause of the bigger difference in other two channels, 
especially in NIR where vegetation cover plays an important 
role. 
     
gray value 
» 
green channel 
    
gray value 
> 
red channel 
    
al 
gray value 
NIR channel 
Figure 4. Channel’s overlap 
86 
Gray values in image channels are strongly influenced by the 
presence of the clouds, since they are a little bit ‘shifted’ 
(lighter) comparing to the clear, non-cloudy areas. 
Creation of the membership functions for the output variables is 
done in the similar manner. Since this is Sugeno-type inference 
(precisely, zero-order Sugeno), constant type of output variable 
fits the best to the given set of outputs (land classes). When the 
variables have been named and the membership functions have 
appropriate shapes and names, everything is ready for writing 
down the rules. 
  
  
  
  
  
  
parameter/output 
class à 
variable 
water ] 
urban 2 
cropl 3 
crop2 4 
vegetation 5 
  
  
  
  
Table 2. Parameter values for output variables 
Based on the descriptions of the input (green, red and NIR 
channels) and output variables (water, urban, cropl, crop2, 
vegetation), the rule statements can be constructed in the Rule 
Editor. 
Rules for image classification procedure in verbose format are 
as follows: 
  
  
IF (GREEN is mfl) AND (REI 
THEN (class is water) 
IF (GREEN 1s mf2) AND (REI 
THEN (class is urban) 
is mfl) AND (NIR is mfl) 
V 
A 
is mf2) AND (^! is mf2) 
IF (GREEN is mf3) AND (REED is mf3) AND ( is mf3) 
THEN (class is crop!) 
IF (GREEN is mf4) AND (RED is mf4) AND ( is mf4) 
THEN (class is crop2) 
IF (GREEN is mf5) AND (RED is mf5) AND ( is mf5) 
THEN (class is vegetation) 
  
  
At this point, the fuzzy inference system has been completely 
defined, in that the variables, membership functions and the 
rules necessary to calculate classes are in place. 
Classification is conducted by the Matlab's m-file. Resulting 
image is showed in the Figure 5. 
Int 
Oui 
fuz 
are 
clas 
urb 
Thi 
pix 
of | 
  
Tat 
Lar 
fou 
Fig