table 6. accuracy analysis with AP’s 
(unit:m) 
without AP's with AP's 
case 
OXy oz | uxy uz | oxy oz | Ixy Oz 
1 7.08/|12.02/10.11/|11.30, 6.22| 9.81,| 8.97. 9.48 
accuracy increasing rate(x) 12.15/18.34/11.28|16.11 
2 16.63|36. 14|38. 29|26.09|16.59|31.48|38.05|20. 62 
accuracy increasing rate(x) 0.2 12.9] 3.2 120.9 
5000 
ANNNNN 
CASE 1 CASE 2 WITHOUT APS 
EE 
4000 ’ WITH APS 
3000 
z 
2000 rd AN NI 
  
  
  
  
  
Fig. 6 accuracy analysis with AP's 
But, for case 2 where accuracy of GCPs are relatively low, 
height accuracy was improved by 10x but for planimetric 
accuracy the improvement was less than 5X due to influence of 
gross errors. High accuracy improvement cannot be expected 
with only AP's in case 2. 
Therefore in cases where gross errors are present in GCPs, 
Simultaneous adjustment method with AP's and gross errors 
elimination process is necessary to compensate for the error 
of GCPs. 
The developed self calibration bundle adjustment program 
includes the significance testing process. The significance 
testing process corrects systematic influence of the AP's for 
the whole adjustment system through the F test which is a 
multi-dimensional test. Then AP's after the multi-dimensional 
test are gone through the one dimensional t test. These 
processes determines the AP's which are significance of the 
level 1AP photogrammetric film and the level 1A digital 
images(Fig. 7). 
  
  
  
  
à NU A 
13.173 
7 
  
Y 
> 
  
  
  
  
  
  
  
  
  
1 
13 14 
  
  
  
  
  
  
  
UUVY VC 
YY tt rt 
  
  
  
A x = azx A x = agy? 
> 
se 
z 810x? 
(a) level 1AP 
376 
  
  
S1] 
VIUA UT UU Y ii 
ET tn AN IN 
tat 
LLC 
TT 
Tt 
1114: 
17771 
11 
/ J 
21 d 
T 
I 
  
ivy 
ILC 
lr 
3.4.1 4 1 
bii: 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
V 
L 
  
  
  
  
A x = 
A x = agy? 
A x = a,sxy? 
(b) level 1A 
Fig. 7 effects of significant AP's determinated from this 
study 
3.3.2 Accuracy Analysis According to Gross Errors 
Elimination 
In the elimination process of gross errors, after confirming 
the presence of gross errors from residuals of image 
coordinates computed by self calibration bundle adjustment of 
cases in table 7, simultaneous adjustment with progressive 
gross error elimination method was applied. 
The determined studentized residual ti and the Baarda test 
Statistics amount ti54.1300 was used which suggests the 
minimum power of test (1-8) to be 80x at 0.1% significance 
level. 
No gross error was detected for case 1, but for case 2 where 
the 
geometrically unstable, gross errors were detected for points 
ground control points and image coordinates are 
no.22 and no.27. The result of simultaneous adjustment of case 
2 is in table 7 and the accuracy improvement due to AP's and 
to gross error elimination is as in Fig. 8. 
table 7 accuracy of 3-D coordinate with elimination of gross 
  
  
  
  
  
  
  
errors 
(unit:m) 
geometric | accuracy of 
precision | check point 
item remarks 
Oxy Oz Uxy Hz 
bundle adjustment|16.63|36.14| 38.29| 26.09 
self calibration |16.59|31.48| 38.05, 20.62 
bundle adjustment AP 
87,815,816 
increasing rates | 0.2 |12.9 3.2 | 20.9 regarded 
of accuracy(x) 
Simultaneous 14.28|26.97| 28.33| 15.30 
adjustment AP 
87,815,816 and 
elimination of 
increasing rates |14.1 |25.4 | 27.9 | 41.4 gross error 
of accuracy(*) NO. 22,27