ets was 
of these 
points, 
irvey. À 
| of the 
> month 
grabbed 
libration 
out the 
second 
relative 
relative 
stem b- 
and (2). 
ters are 
ctor and 
ason for 
es 1s the 
out 3-5 
ustment, 
relative 
sults of 
nera (2) 
obvious 
e three 
relative 
minutes 
lon on à 
s = 3 cm 
lands, a 
cameras, 
, mainly 
"or more 
he error 
City in 
s, urban 
ys with a 
t results 
bout 180 
that the 
system 
e day as 
system's 
| order to 
ks along 
's results 
yackward 
s as the 
he same 
esults of 
Table 5. The relative orientation parameters of calibration (2) without applying the relative orientation constraints 
( Results of Camera (1) and Camera (2) ) 
  
  
  
  
  
  
  
  
  
  
  
Calibration B ^o (deg.) AD (deg.) AK (deg.) 
No. 2 (cm) cam (1)  cam(2) | cam(l)  cam(2) | cam(l)  cam(2) 
Vanset-up 1 | 182.71 | 80.4003 | 81.3991 0.1379 0.7130 -7.8114 9.2847 
Vanset-up 2 | 183.21 | 80.3921 | 81.4491 0.1535 0.7021 -7.8303 9.2422 
Van set-up 3 | 182.44 | 80.4064 | 81.3714 0.1464 0.6200 -7.8164 8.9672 
Van set-up 4 | 183.98 | 80.4453 | 81.3655 0.1849 0.7349 -7.6452 9.2266 
Van set-up 5 | 182.91 | 80.4495 | 81.4095 0.1478 0.6639 -7.7682 9.1123 
  
  
  
  
  
Table 6. The relative orientation parameters of three different calibrations after applying the relative orientation constraints 
( Results of Camera (1) and Camera (2) ) 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Calibration B ^o (deg.) AD (deg.) Ax (deg.) 
(cm) cam(l)  cam(2) | cam(l) cam(2) | cam(l) cam (2) 
No. 1 183.47 | 80.4691 | 81.4191 | 0.1691 0.6313 | -7.8631 9.1923 
No. 2 183.12 | 80.4325 | 81.3991 | 0.1603 0.6778 | -7.8484 | 9.2010 
No. 3 183.33 | 80.4495 | 81.3723 | 0.1642 0.6538 | -7.8521 | 9.1892 
15 Table 9 summarizes the system repeatability as obtained from 
the tests described above. The RMS(x,y) is the composite 
10 4-4. erre 9-8 
T A " ef AZ e horizontal error, computed from: 
8 e oat &449 2. un | 
5 S4 cs... 4A 9$, g^9"95 RMS(E,N) - y(&E)? -(ÓN ? , where 8E and àN are the 
5 0 errors in E and N coordinates. 
= T T T T 
5 d 8 I a 
BS ah Bria hg Ad acu. A 
2 " &a me , s m e A H A ? Repeatability RMS (E, N) RMS (h) 
eO lune user rut tienda Mit S ^ be » (cm) (cm) 
-15 Forward-backward, same day +8 UE 
gan 2nd tone 22 5 12115016 Figs, 2 Same Direction, different. t6 +4 
| eE &8N ah | Point Number days 
3 Forward-backward, different + 10 Hind 
Figure 7. Repeatability of Forward-Backward Runs on the Same davs 
Day f 
15 . . “gr 
e e Table 9: Statistical summary of the system repeatability 
1o. d... 9 3. Boso siii iut 
© A 8 a e . = The results in height indicate that the GPS/INS positioning 
o A $ ; : : . : 
*sleee—————A.. buone acm component is working at the centimeter level. Since the height 
E component in GPS is the weakest, it can be expected that the X 
gm ; ; : and Y components are at least of the same accuracy. The increase 
E in errors for the horizontal components must therefore be due to 
= -5 d us. Nauntheicen eabcoso dea red A the camera array. The most likely explanation is that the increase 
e A ® in RMS(x,y) as compared to RMS(h) is due to the along track 
9, 1.3... 5. : SER CRG uua eer A. AM eu. error caused by the errors in determining the base between the 
e 2 cameras. The across-track error should be small and comparable 
-15 2 in size to the height error. Another explanation would be a poor 
0 2 4 6 8 10 12 synchronization between the GPS/INS component and the 
camera component. This would result again in an along-track 
| eE mN Ah | Point Number 
  
Figure 8. Repeatability of Forward-Backward Runs on Different 
Days 
99 
error. Repeatability in the same direction, even on different days, 
seems to be consistently better than the repeatability between 
forward and backward runs. But the RMS of 10 cm in horizontal 
and 7 cm vertical are still reasonable considering the fact that the 
errors in forward-backward direction should be A times the 
same direction errors. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996