stanbul 2004 
  
r (SIMC) of 
es Landsat-7 
nd reference 
r land cover 
t images are 
ate few steps 
re 4 First 
| parameters 
or minimum 
1 knowledge. 
owing. Scale 
rging means 
ure 6 shows 
are classified 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004 
  
  
  
-elei Band A El a i Bo i ii 
  
  
Me [ues 
Band 1 Band 1 FA 
Band 2 Band 2 SF 
Band 3 Band 3 91 4 
Band 4 Band 4 93 ^l 
Band 5 Band 5 93 2l 
Band 6 Band 6 914 
Band 7 Band 7 SN 
  
-Threshold 
1] Level for Combining : fi 0 
2] Scale : fre 
3] Level for Mergirig : [zo 
  
  
  
  
  
  
  
Barren EH wetland 
  
    
    
  
  
   
   
  
Agriculture 
Grass M Forest 
Figure 7. classification map of (a) large scale as reference (b) this study method (c) pixel based classification 
That is to say, training data is replaced with feature database. 
So, users don’t feel inconvenience to select training data sets. 
Figure 7 (b) shows classification result using feature database. 
Figure 7 (a) is large scale classification map as reference 
produced by ministry of environment, figure 7 (c) is pixel based 
classification map using Earth 2.0 software. Although 
classification result is extracted better using editing classes, we 
did barely work post-processing in view of non-specialists. 
Accuracy assessment is planning in the future in consideration 
of time and area a lot. Examining with the unaided eye, 
accuracy of method in this study is better than that of tradition 
method. It will be expected to serve convenient surroundings to 
Users. 
567 
4. CONCLUSION 
We must select training data in supervised classification. As 
images are classified based on training data, we select training 
sites within image that are representative of the land cover class 
of interest. Users don't feel inconvenience to select training 
data sets sometimes. So, automated classification method using 
feature database is proposed in this study. Feature database has 
statistics calculated training data. We construct statistics about 
brightness, tasselled cap transformation and band ratio in rural 
area, forest area, grass area, agriculture area, wetland area, 
barren area and water area in now. As a result of our developed 
classification software in test area, it is expected that proposed 
method is higher accuracy than traditional method. It will serve 
convenient surroundings to non-specialist users.