luring 
Freiburg, 
7 0.82 um. 
brandt, 
e soil, 
> scene 
ing 
e for 
ned for 
à 
Iss and 
ependance 
real 
es can 
second 
rass, 
tted 
ariances 
ion 
  
  
     
  
  
  
  
  
  
  
  
BARE SOIL 560 TRAININGAREAS WINTERWHEAT 345 TRAININGAREAS GRASS 73 TRAININGAREAS 
o S 
U 
c 
= hed 
z p^. 
e ^ e 
116] ÁN 
100] 
zi 
« 84) I 
Lai . 
u ; 
T T T T T T T Y T T T T T T T T T T 
-&40*  -20* 0° 20° 40° -40° — -20* 0° 20° 40° 
view angle view angle view angle 
o 
= 6 160) 160 
c 
fo 4 - À 
££ 
O48] 120 120 
32] 80] 80] 
“ww 
u 
æ 
x 
a 
<< 
> 
  
  
  
  
  
  
  
  
-40° 20° 0° 20° 40° -40* — -20* 0° 20e 40° „40° -20: l 20* 40° 
view angle view angle view angle 
  
  
  
  
  
  
  
  
  
  
e 
SN 1 
à 7 | zl 
29 80 
o À 
e ; = 
© 40 |: . 40 
= v 
uv 
us 
w 
= 
< 
= - 
< 
> 
© . 
w -80 -80 3 
TIL. Rats Taha (Te Ta IS TW TT DF EAT 
- 40° . -20» 0° 20° 40° - 40°  -20* 0° 20° 40° 40° — -20* 0° 20° 40° 
view angle view angle view angle 
Figure 1 Direction dependant radiance behaviour of bare soil , winterwheat and grass 
from 4000 m altitude. 
(———) fourth order € 
  
) second order and. (— — —) first order polynomials. 
Therefore trendcorrected covariance matrices were calculated in selected scan angle 
ranges. The whole image is subdivided in scan angle ranges of equal size and within each 
range all available training areas are summarized. Then for each range a covariance 
matrix is calculated, correcting the trend with a polynomial to determine the mean for 
this scan angle. Figure 2 show trendcorrected variances and covariances, determined by 
subdividing the strip into 10 scan angle ranges. Second order polynomials through data 
points are also plotted. The graphs indicate a change of covariance matrices with scan 
angle which can be different for each object and wavelength.