acy 
Xy 
  
.150 
2097 
„152 
„093 
:155 
  
  
it GPS 
with 
etter 
it are 
5, in 
; were 
= the 
> same 
uracy 
cates 
- case 
| the 
sults 
n all 
ta. 
the 
| were 
five 
cally 
lling 
e use 
ction 
inear 
S per 
4 
  
zd 
> 
  
| — I 
i 
b 
  
EL 
cint 
nt 
leters 
the 
ie GPS 
inear 
were 
chosen (Cl, |03..and ..€5), to check --if ‘the 
significance of the correction parameters depended 
upon the control configuration. 
The tests were carried out per parameter group, 
and the results are given in table 4. 
  
  
  
  
  
  
  
configuration |strip| critical value test 
no. quantity 
F 
3,9,.003 CT LT QT 
1 292.0 184.7 0.5 
C1 2 43.0 116.0 0.5 
3 10.8 63.5 0.9 
4 29.0 49.4 0.3 
1 549.0 6.0 0.9 
C3 2 4.7 527.0 121.0 0.5 
3 664.0 15.6 0.1 
4 6.0 8.0 0.1 
1 300.0 31.6 0.1 
C5 2 726.4 21.9 0-3 
3 791.0 24.2 0.5 
4 6.7 21.9 0.3 
  
  
  
  
  
  
  
  
Table 4. Test quantities for the groups of 
parameters. 
From the above table we see that the test 
quantities for both linear and quadratic terms are 
rejected for all strips in all configurations. 
This indicates that these two terms are 
significant. On the other hand, the test 
quantities for the quadratic terms are smaller 
than the critical value, which indicates that the 
quadratic terms are not significant. 
The conclusion does not necessarily imply that the 
quadratic deformation is not present in the GPS 
data. As we have seen in the experiments on the 
statistical significance with generated data, the 
quadratic deformation can be very well 
approximated by the linear terms used in the 
adjustment. It is also observed that the results 
are consistent and independent of the 
configuration used. 
Based on these outcomes, the proper modelling for 
the combined adjustment seems to be the one which 
takes into account the constant and linear terms. 
The introduction of the quadratic term may 
negatively effect the results. 
The GPS modelling vas done per strip, and it was 
of interest to assess if the parameters introduced 
per strip were significantly different. The 
individual parameters of strip 1 and strip 2 in 
configuration C3 were tested. The critical value 
for this test was Fi wi. 001. 10.8. Since it had 
been already proved that the quadratic term was 
not significant, the test among parameters was 
performed for only the constant and linear terms. 
The results are given in table 5. 
  
  
  
  
  
  
  
  
GPS modelling |parameter diff |test quantity 
term in 
X 29.0 
constant y 89.3 
z 12.3 
X 25.2 
linear y 26.7 
zZ 16.5 
Table 5. Test quantities between individual 
parameters 
507 
Ve can observe that the differences of the 
parameters are significant. Similar results vere 
obtained by testing the significance of the 
parameter differences among the other strips. This 
suggests that modelling of the GPS parameters as 
strip invariant is appropriate for the available 
set of data used in the experiments. 
Table 6 contains the accuracy results of the five 
configurations shown in figure 2, being adjusted 
as follows: 
- Without GPS data 
- with GPS modelled with linear and constant terms 
- with GPS modelled with linear, constant and 
quadratic terms 
  
  
  
  
  
No. of rel. accuracy |abs. accuracy 
Case check |config. (meter) (meter) 
points 
c o c u u u H 
x y z x y z xy 
23 C1 -017|.019|.032|.090|.070|.533|.081 
without 19 C2 .017|.019/.032|.108].043|.418|.092 
GPS 13 C3 -020|.020|.033|.097|.042|.109|.074 
data 13 C4 -020|.020|.033|.039|.042|.108|.041 
24 C5 -018|.019|.032|.052|.037|.255|.045 
vith GPS | 23 C1 -019|.020|.033|.092|.054|.429!.074 
data 19 C2 -020|.020|.034|.120|.047|.130/.091 
(constant 13 C3 -021|.020|.034{/.097|.051/.108|.078 
& linear | 13 C4 .021/.020|.035|.035/.030/|.098/|.033 
term) 24 C5 .020/.020|.033|.068|.036/|.139|.054 
with GPS | 23 C1 -018/.019|.032|.091|.072|.598|.082 
data (con-| 19 C2 -019|.020|.033|.121|.047|.434|.091 
stant lin-| 13 C3 -020|.020|.033|.118|.042|.106|.088 
aer & qua.| 13 C4 .021|.020|.034|.030/|.035]|.103|.037 
terms) 24 C5 .019|.019|.033/.055|.032|.259|.045 
  
  
  
  
  
  
  
  
  
  
  
  
Table 6. Variation in control point configuration 
and in GPS modelling. 
Comparing the results of the adjustment "without 
GPS data" with "GPS constant and linear terms", it 
can be observed that the planimetric precision is 
approximately the same, while the height precision 
shows considerable improvement. The highest 
improvement, by a factor of 3.2, is observed for 
configuration C2. 
Comparison of the results shows that with the 
introduction of the quadratic term the height 
accuracies deteriorate quite appreciably for 
configurations C1,C2 and C5, while the accuracies 
remain approximately the same for configurations 
C3 and C4 because these configurations are very 
well controlled in height. 
Comparing the results of the different control 
configurations in the combined adjustment with 
constant and linear terms for the GPS modelling, 
we see that 
- One additional control point in the middle of 
the block configuration C5 compared with C1 
improves the height precision by a factor of 4. 
- Configuration C2 gives height precision similar 
to C5. 
- The height accuracy of configuration C2 is only 
2 cm more than the accuracy of C3, but C2 has 
only two chains of height control points, while 
C3 has 3 chains. 
CONCLUSIONS 
The results demonstrate that GPS-controlled 
photogrammetry has the potential to reduce 
substantially the need for geodetic control points 
while the accuracy requirements are maintained. 
The improvement in height accuracy is significant: 
improvement by as much as a factor of 3.2 has been 
observed in the experiments with the real data.