'ssed using the
| independently
This software
rch techniques
olution of the
bh significantly
ns which must
ection software
cult portions of
| TRIAL
ed Nottingham
trial. By taking
ilst moving the
calculated for
proach to that
| and widely
n also enabled
the timer chip
ses the results
observed twice
) dL (m)
0.076
) 0.075
) 0.058
| 0.049
»pancies.
n of the vector
approximately
fected by the
1 easily vary at
ic accuracy is
also a factor that must be taken into account, particularly
as some of the targets were badly blurred.
Additionally, these analysed frames were over exposed
due to a fault in the internal mechanism of the UMK.
Knowledge of this factor possibly explains the position
discrepancy of 0.9cm between frames 3 1a and 3 1b.
To explain the difference between the vector offset
accuracy (dL) for the two separate frames it is necessary
to look at the standard errors of the NOTF software L1
carrier phase solution. It can be seen that the standard
errors are higher for x,y and z in the case of frame 4 4,
corresponding to the greater offset vector. In good
geometric conditions, NOTF has been shown to
comfortably resolve the ambiguities for positioning at the
2cm level. Unfortunately, the satellite geometry during the
evaluation trial was adversely affected by the building
facade.
| Frame | ox (m) | ey (m) | ez tm ]
| 31 [| 0.0397 | 0.0197 | 0.0348 |
| 44 | 0.0460 | 0.0250 | 0.0439 |
Table 2 - Standard Errors for NOTF L1 Carrier Phase
Solution at Exposure Station.
In summary it is clear that the new integrated system
performs significantly better than the previous one. The
evaluation trial was not perfect in that the photography
was over-exposed due to a fault in the internal
mechanism, and the satellite geometry was obscured by
the height of the building facade. However, the trial
objectives had still been met successfully.
5. COMBINED ESTIMATION OF SIMULATED GPS
DATA IN A REAL FLIGHT TRIAL
As part of the planning of a final flight trial, it was felt that
some combined block estimation work with simulated
GPS data should be undertaken. The photographic data
was from the 15 frame initial evaluation trial undertaken in
March 1994 (Smith and Joy, 1995a). All measured points
were targets to represent typical control points. Targets
were also used at the minor control points since the field
surface was ‘soft’. The raw GPS data was simulated from
the positions provided from a fully controlled
photogrammetric block, with information on the satellite
geometry provided from the IGS precise ephemeris.
Configuration Unknown Point Standard Errors (mm)
Centre 3.0 3.0 4.1
Centre+Left 3.0 2.9 3.8
Centre+Right 3.2 3.0 4.3
Full Block 2.9 2.9 3.7
Conventional kinematic OTF processing was then used to
obtain exposure station coordinates.
The data processing was performed with the in-house
software TABBY (its name does not serve as an
acronym) running on an SGI Indigo Workstation. TABBY
is a combined GPS Bundle Estimation program written by
the author during the period of research and has
undergone a suitable level of testing prior to operational
use. The familiar collinearity equations are used
alongside observation equations for the antenna-phase
centre offset vector. Control points are introduced as
weighted observations because of the precision required
by the application. For this adjustment work, the vector
offset was held fixed in magnitude in the observation
equations and no self-calibration or drift parameters were
estimated. The primary aim was to quantify the effect of
adding phase centre observations into a bundle
estimation in the simplest manner possible.
A Priori estimates for the control standard errors were
entered as +0.005m in plan and +0.002m height. The a
priori image coordinate standard error was 3um. Table 3
shows positional standard errors for both the control and
the unknown points in the photographic test field resulting
from processing of several strip combinations. No GPS
positions were used, just as in a traditional
photogrammetric block. These results are an update from
those in Smith and Joy (1995a) and show that the control
point standard errors are significantly improved in plan
and marginally in height for all cases.
It can be seen that there is an improvement in precision
by the addition of photographic strips, with control points
consistently of a higher precision to the unknown points.
The Centre+Right block is poor because of blurring
dictating observation of fewer points. In further
discussions this configuration will be ignored as the
Centre+Left block is more representative of the precision
increase that would be expected.
GPS antenna coordinates were then entered in the
estimation under 4 distinct control configurations. These
were GPS with no ground control, GPS with 4 full 3D
control points in the corners of the block, GPS with the
same 4 points plus 6 height points, GPS with the full
control configuration, and the values from Table 3 (no
GPS used) for comparison. The average standard error
for the antenna phase centre positions was 0.5cm plan
and 0.7cm height (discussed at the end of this section).
Control Point Standard Errors (mm)
2:6 2.6 1.8
2.6 2.6 1.8
2.6 27 1.8
2.5 2.6 1.8
Table 3 - Conventional Bundle Estimation On Pontefract Test Field.
817
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
