with nearest 
'rgent camera 
Optical axis, 
the effects of 
g, an affine 
ree reference 
are projected 
equal height 
in the search 
nd unshaped 
e two corner 
patch. Affine 
ed from the 
haped search 
je parameters 
yrmed grid at 
been used in 
position. 
lera geometry 
lel to the XY 
stant Z-value 
ons. If any of 
n heights, for 
1 be used to 
surface patch 
is which have 
[C search the 
an be used as 
convergence. 
reliability of 
ional stereo- 
ning a highly 
D camera in 
aphs of the 
ed in figure 3 
average for a 
ce image (no. 
nera positions 
it 45 degrees 
shaping. The 
h shaping and 
the maximum 
d the average 
rom the patch 
ge correlation 
) than for the 
orrespond to 
eir maximum 
f graphs also 
haping which 
age 6 and thus 
ise of multiple 
natching even 
arameters still 
  
p (2) 
-05 | 
    
   
-0.8 1 
else 
94 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
  
  
0.8 
0.6 
0.4 
0.2 
p (6) 
=0.2}- 
-0.4 
-0.6 
-0.8 
N 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
Z 
  
  
0.8 
0.6 
0.4 | 
0.2 | 
p (4) 
-0.2 
0.4 +— 
-0.6 
-0.8 
  
=) 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
Z 
  
  
  
p (mean 2, 6, 4) 
  
  
  
  
  
Figure 3a. MIC values of test search 
using images I, 2, 6 and 4 with patch 
shaping. 
5. SURFACE MEASUREMENTS 
Two examples are shown here, the first is the measurement of 
the blades of a propeller (see figure 4) and the second of a 
smooth curved plastic lid (see figure 5). As was to be expected 
the MIC procedure was called more often for the propeller 
measurement than for the smooth curved surface. The results 
from the MPGC matching are shown in table 1. The poorer 
precision in the depth coordinate (Z) is due to the fairly narrow 
camera base between the images employed in these 
measurements. A more convergent imaging geometry would 
increase the precision in the resulting Z coordinates. The 
resulting surface data was exported to a surface presentation 
503 
  
P (2) 
  
= 4 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
  
  
  
1 
0.8 
0.6 
0.4 
0.2 
p (6) 
0 
-0.2 
-0.4 
-0.6 
-0.8 
  
  
  
-1 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
  
  
1 
0,8 31— — 
0.6 
0.4 
0.2 
p (4) 
0 
-0.2 
70.4 
-0.6 
-0.8 
- 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
  
  
a 
QB1—- 
0.6 
0.4 
0.2 
0 
mean 2, 6, 4) 
-0.2 
SZ 04 
a 
-0.6 
-0.8 
=} 
940 960 980 1000 1020 1040 1060 1080 1100 1120 1140 
  
  
  
Figure 3b. MIC values of test search 
using images 1, 2, 6 and 4 without 
patch shaping. 
package called SURFER and the perspective plots of the two 
surfaces are shown in figures 6 and 7. 
6. CONCLUSIONS 
The matching procedure presented here is flexible to the type of 
targeting used. In this work projected grids and hand-drawn lines 
have been employed for providing surface detail. Natural texture 
could also be used. The only restriction is that the surface detail 
is fairly dense and of sufficient contrast for matching. 
A matching procedure has been designed, based on the MIC 
search procedure developed in this work and on the existing 
MPGC matching method. Some points concerning the MIC 
algorithm are listed below. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996