en performed on those 
1atched node relations 
wn. Since the node 
there were no wrong 
equation (1), the base 
uted. The processing 
seconds for the OSU 
ion only is shown in 
‚The matching result 
own in Table 3, two 
rilarly except for the 
> search, it tries to find 
inimum number of nil 
recognizes the cost of 
e solution path already 
btrees as long as the 
ler than the path that is 
or performance of the 
a path which has a 
| smaller number of nil 
e path as the solution 
efficiently with the nil 
on suffers from the 
necessary subtrees. 
nction does not have a 
oint of computation 
enefit function finds a 
solution path already 
th without considering 
nt path. This property 
h tree backtrack when 
benefit of the solution 
Function leads to better 
luation functions has 
€ held only locally, but 
re obtained when unit 
primitives in a set of 
s which happen to have 
relations among the 
ied line primitives, the 
s was computed. The 
the matched lines and 
the mismatch detection. 
int numbers after the 
tection were used for 
en two images. Table 5 
he proposed relational 
  
  
  
  
  
  
  
ost Benefit 
ction Function 
97 197 
03 107 
0 97 
13 10 
379 90.654 
128 1079 
  
  
  
na 1996 
  
  
  
  
  
  
  
% 
  
  
after 
feature 
(b) (b) 
Figure 6: OSU left (a) and right (b) images superimposed with Figure 7: OSU left (a) and right (b) images with extracted 
interest corner points: 9796 confidence using 7x7 corner points and straight lines 
operator window size, 7x7 non-maxima window size, postprocessing. 
q — 0.75, and f 1.5. 
117 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996