aluation 
> Means 
ification 
ad error 
nd non- 
  
used for 
data, b) 
[aximum 
error at 
relation 
he result 
the tree 
  
  
  
  
  
  
  
  
Cluster | descriptio | ground merged merged Fuzzy C- | merged merged 
n truth data | classes I classes I | M. classes I classes II 
area % area % area % area % area % area % 
1, Fig.8 | grass 58.34 58.34 41.02 57.02 57.02 
Aral low veget. included included 16.00 included included 
in 1 in 1 in 1 in 1 
2s road 41.66 41.66 29.74 42.98 42.98 
qom concrete included included 13.24 included included 
in 2 in 2 in 2 in 2 
1, Fig.9 | bush 18.73 included 22.30 included | included 
in 3 in 3 in 3 
3e forest 43.94 62.67 48.54 70.84 70.84 
2, grass 37.33 37.33 29.16 29.16 
  
  
  
  
  
  
  
  
  
  
  
  
5. CONCLUSION 
The results of the analysis show that the announced images 
of the high resolution satellite Quick Bird would give a 
significant contribution to the studies of forestry, vegetation 
and human settlements compared with the satellite images 
of Landsat TM, Spot and others. A 4m resolution images 
with bands: Red, Green, Blue, Near IR, sharpened and 
merged by Im panchromatic image have turned to be the 
best of all the analysed images of 1m,3m, 4m,15m 
resolution. Of 6 analysed methods, Fuzzy-C Means and 
Maximum Likelihood method as supervised method has 
proved to be the most efficient. Natural areas are classified 
with the error up to 8% . Likely source of errors are shadows 
at isolated trees, shadows at clearance in the forest as well 
as resolution not being precise enough to image the tree top. 
Shadow is the biggest source of errors particularly 
pronounced with trees and buildings. At several houses 
where shadows can be well detected it was found that error 
at measuring 1m ground-truth area can be 28-49%. With 
human settlements, increase in classification evaluation is 
likely to be expected if division is made within the 
settlements themselves into more classes excluding the 
shadow areas. 
6. ACKNOWLEDGEMENTS 
It is with great appreciation that I thank to Earth Watch for 
their imagery and to Microlmages for their software. 
7. REFERENCES 
Earth Watch, 1996, Digital Globe sample imagery, CD, 
Version 1.0, Earth Watch, Longmont, CO. USA. 
Fritz L.W., 1996, Commercial Earth observation satellites, 
International Archives of Photogrammetry and Remote 
Sensing, Vienna, Vol. XXXI, Part B4, pp.273-282. 
Microlmages, 1997a, Announcing TNTlite, Memo, 
Microlmages, Lincoln, Nebraska, USA. 
Microlmages, 1997b, TNT Products V5.6, CD, 
Microlmages, Lincoln, Nebraska, USA. 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 661