i 
[ 
H 
t 
1 
i 
1 
  
    
     
  
   
     
     
    
     
    
  
  
   
    
    
   
       
     
   
— IT To 
USICNS ALD RECOLXELDATICES : 
2 
C 
ca 
C 
tt 
C3 
r 
  
The bivariate smoothing process presented in this paper proved 
to be efficient end Seopemis + The idea of employing only one 
parameter to change the smoothing degree makes the method 
flexible and easy to control. The use of the moving coordinate 
System results in small storage capacity | which is independent 
on the block size or the number of points. : 
The analysis of the results of the Transfer Function program 
snows linear relationship between the norm E and the degree 
of smoothing at each frequency, this relationship is valid for 
all terrain types. Similar relationship also exists between 
the residual function € and the parameter W, at constant 
frequency,l Fig.5 y+ Which means that for a civenTterrain type, 
137 
the increase in the value of Wy reduces e linearly, FPiz.6 . 
- 
These relationships facilitate the pre-determination of the 
smoothing dceres which corresponds to a specified value of 
the residual function. 
"hen &pplying the presented smoothing process on real DEM 
blocks, oversmoothing may occure in rough eresns Fig. this 
over smoothing can be controlled to filter out the extreme 
values which may be regarded as gross errors. However, tnis 
filtering effect is not unlimited as can be seen from figure 4. 
The maximum smoothing that can be achieved is equal to(1-W, ) 
which is independent on the terrain type. 
It is recommended to smooth DEM data before usin ng them in 
& collocation surface fitting technique. The maximum smoothing 
degree is determined according to terrain type and the amount 
of noise expected in the data. 
l. Lafera,R.1L.,1973,"Computer Methods for Science and Engine- 
ering",Intertext,USA. 
2. perarevt. D..,1970,"A Dizital Terrain Model System", ITC 
ournal-1,p:57-82. 
3. Scheid,T.,1968,"Th eory and problems of numerical analysis", 
Schaum's outline SPAN Hill. 
^ ^T mo 
and 9 
10 +