2004 
y the 
ifted’ 
les is 
rence 
iable 
n the 
have 
iting 
NIR 
rop2, 
Rule 
t are 
  
mfl) 
mf2) 
mf3) 
mf4) 
mf5) 
  
etely 
| the 
Iting 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B6. Istanbul 2004 
  
Figure 5. Classified SPOT images (fuzzy classifier) 
Output images coming from PCI maximum likelihood and 
fuzzy classification can be compared. These grayscale images 
are produced in such way that pixels coming from the same 
class have the same digital numbers in both images: water (50), 
urban (100), crop 1 (150), crop 2 (200) and vegetation (250). 
This is the basis for image comparison. Percentage of classified 
pixels in both methods is given in the Table 3 (overall number 
of pixels 1s 10743070). 
  
  
  
  
  
  
  
  
  
  
  
method PCI fuzzy difference 
class 
water 1.25 1.39 0.14 
urban 15.62 13.95 1.67 
crop 1 13.1 17.24 4.14 
crop 2 28.82 34.11 5.29 
vegetation 37.90 29.99 7.91 
  
Table 3. Percentage of classified pixels in ML and fuzzy 
classification 
Large number of misclassified pixels (black pixels) can be 
found in the areas covered by clouds (yellow circle regions in 
Figure 6). 
87 
Figure 6. ML and fuzzy classification comparison image 
3.4 Accuracy assessment 
Idea for accuracy assessment of fuzzy logic classification 
results. comes from the manner the maximum likelihood 
accuracy assessment was performed: select random sample 
areas with known classes and then let fuzzy logic ‘say’ what 
these samples are. With 100 random selected samples, results 
were as following: 
= correctly classified samples: 89 
=» misclassified: 11 
=» accuracy: 89% 
3.5 Concluding remarks 
Considering chosen land cover classes, results from image 
classification (Figure 5) and accuracy assessment can be good 
starting point for certain analysis: 
=» in the knowledge base, it must be well known whether 
selected sample is. vegetation (forested area) or 
vegetated crop area 
=» around 30% of misclassified samples represent classes 
with small signature separability 
=» classification procedure is strongly influenced by the 
presence of clouds. These regions are lighter, so they 
cannot be properly classified. Since several samples, 
during accuracy assessment, were taken in this area 
with intention, overall classification procedure is 
probably of higher accuracy 
=» at first sight, time necessary for fuzzy classification is 
longer comparing to maximum likelihood procedure, 
which takes several seconds to classify an image. But, 
if in ML procedure possible image transfer to 
recognizable format for certain software, formulation of 
the training areas, analysis concerning signature 
separability take place, than situation is quite different: 
fuzzy logic takes advantage of already created simple 
rules and image classification (started from the 
scratch in both procedures) equal or even less time 
consuming. Of course, different conditions during 
image capture must be taken into account. 
=» considering the level of classification accuracy, fuzzy 
logic can be satisfactory used for image 
classification.