ul 2004
data,
r EO
lated
EO
vn in
OS
n that
|. mis-
L has
1ce in
ilated
T
A/QC
S the
rated
data,
lation
S/INS
icwed
n the
ISAT
at no
focus
| with
data,
ure |
r von
asting
orthing:
ertical
ted
Figure | clearly indicate that assisted triangulation using | strip
of simulated data can improve the check point RMS, but it is
unable to reach a similar performance of the POS AV 510
Direct Georeferencing data, especially in the vertical. However
it is still acceptable for some lower accuracy requirement
projects. Another conclusion that can be drawn from Figure 1 is
that an increase in PPVG value can not improve the result of the
Integrated Sensor Orientation. Thus, using more tie points in
assisted triangulation will actually degrade the performance. To
investigate this further, additional strips (2, 3 and 4) were used
in the test. Figure 2 presents the ISO test when using 2 strips of
simulated POS AV 310 data.
Integrated Sensor Orientation using 2 Strips of Simulated Data
0.3
2
N
o
: © Easting |
| Northing|
:|D Vertical
Check Point RMS (m)
e
=
o
o
a
2 3 4 5 6 7 8 9
Point Per Von Gruber
Figure 2. ISO Test Result when using 2 Strips of Simulated
POS AV 310 Data
From the above Figure, further improvement in both horizontal
and vertical has been achieved. In fact the final accuracy
achieved by running the Integrated Sensor Orientation on only
2 strips of the simulated data is similar to the Direct
Georeferencing performance of the POS AV 510 system (Table
6.0), and increasing the number of strips in the adjustment does
not significantly improve the results. Therefore the minimum
requirement for assisted triangulation is only 2 strips.
Regardless of the number of strips however, the more tie point
that are used, in the assisted AT, the poorer the results are.
l'urther investigation on this behaviour will be given later.
5.2 SO Testing on POS AV 510 Data
Running Integrated Sensor Orientation on the POS AV 510
system will provide an insight into how much an improvement
that can. be achieved when using a high end Direct
Georeferencing system. The test configuration is the same as to
run the test using the simulated data. Starting with using 1 strip
of the POS AV 510 data, the result is shown in Figure 3.
Integrated Sensor Orientation using 1 Strip of POS/AV 510
Data
: iD Vertical |
%
= 02 -
X Soy
€ : @ Easting |
S 0.15 8 Northing|
x
o
o
4A
o
Point Per Von Gruber
Figure 3. ISO Test Result when using | Strip of POS AV 510
Data
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part BS. Istanbul 2004
From Figure 3, it can be seen that running assisted triangulation
using | strip of POS AV 510 data did not make any
improvement in the achievable final 3D position accuracy
(Table 6). Such triangulation in fact degrades the performance
because of the lack of information in resolving the Omega angle.
Solving this will requires a lot of ground control point in the
strips, which is not practical for many applications. This
explains why a high end DG system is required for single pass
corridor flights, for example. In comparison, Figure 4 presents
the result when 2 strips of POS AV 510 data is used.
Integrated $ensor Orientation using 2 Strips of POS/AV 510
Data
o
N
e
i
{
i
i
i
i
o
=
n
= 02 Re
œ
E Gi Easting
g 5 & Northing
x 10 Vertical
o
D
=
oO
o
o
an
Point Per Von Gruber
Figure 4. ISO test result using 2 Strips of POS AV 510 Data
Even these results clearly indicates that only slight
improvement can be achieved when running assisted
triangulation on the POS AV 510 data. This can be explained
by the fact that having a scale of 1:6000 and using a high end
Direct Georeferencing system, this data set already has highly
accurate 3D position accuracy, and it is especially well
calibrated system, with boresight and datum shift error being
minimized. Therefore, by applying Integrated Sensor
Orientation on this dataset, insignificant improvement will be
made. Notice that the test is performed without any ground
control points. In a case where datum shift might exists in the
project, running Integrated Sensor Orientation with 1-2 ground
control points would improve the result.
Similar to the simulated data, the ISO test on POS AV 510 data
shows a trend that by including more automatic collected tie
points into the assisted triangulation, the results are slightly
degraded. This can be explained by the fact that by using more
tie point in assisted triangulation will create more noise in the
adjustment, as there is no guarantee that the tie point collection
module can maintain the same level of accuracy all the time.
This has been proved when reviewing the parallax for the tie
and the check points through EO Analysis using the assisted
triangulated EO derived parameters. Table 10 lists the parallax
result from the ISO test on 4 strips of simulated data.
PPVG Point Parallax (um) Model Parallax (um)
Value | Max Value RMS Max Value RMS
2 23.2 S2 27 4.3
4 27.3 5.5 7.5 4.4
6 28.3 5.4 6.3 4.2
8 29.5 5.4 6 4.2
Table 10. Parallax Result on EO Analysis after running
Assisted-Triangulation on 4 Strips of Simulated POS AV 310
Data