ibul 2004 
  
  
  
  
  
ormed to 
f spectral 
ata only), 
sification 
raining 2 
yield P3, 
idditional 
products. 
between 
accuracy 
uracy of 
rror) and 
re tested 
roduct 1, 
the error 
/o greater 
ch single 
ement in 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B-YF. Istanbul 2004 
470 000 472 000 474 000 476 000 
4 480 000 
—4 478 000 
-4 476 000 
1.4474 000 
  
— 4472 000 
  
  
  
LEGEND 
[CT] Auicultuie 
Range-shrub 
LI] Range-heib 0 1000 
N 
2000 meters A 
M Forest 
Figure 5: Product 1: Classification Product of bands used as input 
and trained by T1 (Training set for bands) 
470 000 472 000 474 000 476 000 
       
»—4 480 000 
-4 478 000 
  
  
4472 000 
E. 
  
  
LEGEND 
CO Agriculture 
Range-shrub 
[7] Rangehei. 0 1000 
N 
2000 meters A 
ME Forest 
Figure 6: Product 2: Classification Product of bands, DTM and 
slope used as input and trained by T2 (Training set for bands, DTM and 
Slope) 
57 
  
  
Class GT 1 GT2 GT 3 GT4 Total. Accuracy 
1 120 41 34 3 198 60.6% 
2 97 514 125 11 747 68.8% 
3 170 265 710 3 1148 61.8% 
4 0 44 0 43 87 50.5% 
total 387 864 869 60 2180 
accuracy 31.0% 64.5% 81.7% 0% 
Overall Accuracy 63.6% Khat Statistic 41.6% 
  
  
Table 2: Error matrix for Productl; (1) Agriculture, (2) range-shrub, (3) 
range-herb., (4) Forest 
  
  
  
Class GT I GT2 GT3  GT4 Total Accuracy 
1 187 19 32 1 239 78.2% 
2 25 646 107 14 792 81.5% 
3 175 730 730 2 1070 68.2% 
4 0 0 0 43 79 54.4% 
total 387 864 869 60 2180 
accuracy 48.3% 74.7% 84.0% 71% 
Overall Accuracy 73.6% Khat Statistic 58.8% 
  
  
  
Table 3: Error matrix for Product 2; (1) Agriculture, (2) range-shrub, (3) 
range-herb., (4) Forest 
  
  
Class GT GT23 GT 3 GT4 Total Accuracy 
I 127 35 33 2 197 64.4% 
2 9] 520 117 11 739 70.4% 
3 169 265 719 4 1137 62.1% 
4 0 44 0 43 87 50.5% 
total 387 864 869 60 2180 
accuracy 32.8% 65.3% 82.7% 0% 
Overall Accuracy 64.7% Khat Statistic 43.2% 
  
  
  
Table 4: Error matrix for Product 3; (1) Agriculture, (2) range-shrub, (3) 
range-herb., (4) Forest 
CONCLUSION 
In this study, a method primarily based on integrating ancillary 
data into classification procedure as a component is presented. 
The results of the classification with the integration of 
topographical data verified that the method yielded a reasonable 
amount of improvement in classification where conventional 
logical channel approach provided only slight amount of increase 
in total accuracy. 
Highest improvement is obtained for agriculture and lowest for 
forest. Classes, when put into sequential order to comprehend 
relative improvement due to integration of topography show the 
same sequence with the classes listed sequentially by means of 
their correlation with topographical parameters. The case presents