Table 1. Landcover Categories in the small study sites on Ground Truth Map k: 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
(Unit:%) NK 
A(k 
Study area-l | area-21 | area-22 | area-31 | area-32 | area-33 | area-5 sit 
Sites Bid 
sit 
Paddy 0 20.6 0 1.2 1.0 1.2 0 
Forest 32.7 69.1 91.8 37.2 47.7 68.8 44.6 an 
Upland 5551 3.1 3.8 42.7 30.0 5.5 30.4 act 
si 
Waste 0 3.8 0 5.3 3.3 0 6.3 la! 
Water 0 0 0 0 0 0 0 by 
es: 
Residence 1l. 343 4.4 13.6 12.9 24.5 18.7 es 
Ground 0 0 0 0 5,1 0 0 
Total 100.0 100.0 100.0 100.0 100.0 100.0 .1- 100.0 
th 
Table 2. Estimation Root Mean Square Errors of Landcovers pr 
in study sites (Unit:$) qu 
Study | area-1 | area-21 | area-22 | area-31 | area-32 | area-33 | area-5 of 
Sites li 
to 
RMSE 5.4 5.2 3.5 5,7 4.6 1.8 5.0 t 
N(k) 2 
>, (A(K,i) — Ar(k, 1) fe 
ie 1) 
RMSE(k) = - 
us IN (Kk) us 
in 
k: Study Site ID fo 
N(k): Number of landcovers in the study site (k) 2 
A(k,i): Estimated area ratio of landcover(i) in the study site(k) ) 
by the decomposition of mixed data es 
Ar(k,i):Actual area ratio of landcover(i) in the study site(k) de 
3) 
la 
Table 3. Effect of introducing the estimated area ratios of Ba 
landcovers as a priori probabilities in study sites of 
(Unit:$) li 
4) 
Study area-l | area-21 | area-22 | area-31 | area-32 | area-33 | area-5 es 
Sites St 
RMSE-A 7.075 2.10 2.48 6.93 7.78 7.40 6.22 tr 
5) 
RMSE-B 2.45 2.24 0.92 3.36 3.37 1.58 2.54 fc 
si 
RMSE-A: Maximum Likelihood Classifier with even a prioprobabilities of landcovers 6) 
RMSE-B: Bayes' Classifier with estimated a priori probabilities of landcovers tr 
la 
4.Effect of introducing the estimated area classification errors were evaluated by the 7) 
ratios of landcovers as RMSE as shown in the following equation. us 
a priori probabilities ne AV 
The effect of introducing the estimated area : > AN, 
! RE is X, (Alk, i) — Ar(k, i)) 
ratios of landcovers as a priori probabilities ; Se 
i=1 
for the Bayes' classifier is shown in Table 3. RMSE(k) Em N(k) Je 
The digitized ground truth map was compared ir 
with the classification results. The ac 
Tc 
996 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996