ined 
lumes 
Station 
oectral 
id soil 
le 
te. 
ilyzed 
ibil- 
ilted 
shown 
- 1783 - 
On the basis of the analysis, classes 7 and 9 were removed from 
further classification. The results, at this stage, were not 
encouraging with respect to soil classification. 
The results of the supervised classification of timber types are 
summarized in the matrix of Table 3, which shows the performance 
pixel by pixel. 
TABLE 3. The pixel by pixel performance matrix of the timber 
type classification (Huittinen test site). 
  
  
MN Ground Spruce. Pine Mixed Hard Other Total 
> conifer wood 
Landsat 
classifi- $ 
cation 
Spruce 85 8 7 - = 100 
Pine 6 27 7 - 10 100 
Mixed conifer 28 32 40 - - 100 
Hard wood - 6 18 52 24 100 
  
  
Another way of evaluating the results is to look at the perfor- 
mance by timber type units or stands. The classification has been 
compared with field map stands, and the results are shown in Table 
4. If more than 2/3 of the pixels inside a fieldmapped stand 
belong to the same class the classification is accepted as correct. 
TABLE 4. The stand by stand performance of the timber type classi- 
fication (Huittinen test site). 
  
     
  
  
  
  
Spruce Pine Mixed Hard Peat- Seed- Clear- Total 
coni- wood land lings cut no of 
Land- fer stands | 
sat | 
classi- | 
fication | 
Spruce stand 24 5 5 à 35 ln 
Pine stand | 5 14 3 5 3 30 
Mixed conifer 1 6 2 9 
| 
Hard wood 
stand i 1 12 
Outcrop or 
peatland | i 14 5 20 
! 
Peatland | 
(pine) | 3 10 16 2 31 
Clear cut | 4 8 12 
| 
Misc. (<2/3) 6 6 5 1 8 
  
Total | 37 42 20 il 23 24 10 167