iver. When
lemands for
a release, it
re the given
large scale
) DGPS
TION OF
PS is used
the camera
s, due to a
points have
st-processed
\Ithough, the
liminated by
s of modern
1 centimeter
achieve this
arrier phase
| the phase
termine the
) exploit the
ral methods
juities in an
l| be briefly
/S occur, the
values. In
2d in a static
continuous
1 flight turns
t losses of
frequent in
OF), tries to
om the GPS
sts are used
nd incorrect
djustment of
(CBA). The
| two step
les are fixed
step and the
ective center
stment using
nd the post-
8 [1990],
importance,
unfavorable
e flying time
sing between
> conditions
' reached an
1990], Hatch
ed in AROF
are based on the assumption, that the carrier phase
observations are unbiased, therefore several side
conditions have to be observed when trying to do AROF:
e The distance between the reference station and the
roving receiver may not exceed 10-20 km so that ail
common systematic error effects are canceled out
when differencing the observations from the
reference station and the rover.
e At least 5-10 minutes of continuous data is required
for a successful ambiguity resolution, because a
certain number of observations are required to
achieve a maximum significance level in the
statistical tests.
e No, larger biases may be on the phase observations
(e.g. from multipath or larger tropospheric
differences) because otherwise the statistical tests
produce incorrect results.
e Dual frequency receivers are required, to allow for
widelaning observations
Until today, these prerequisites have prevented a
successful and economical use of AROF in airborne
photogrammetric applications. (see e.g. Schade [1992]).
The reasons speaking against AROF in an airborne
environment are:
e the critical logistics: it is often difficult to have a
reference station within a radius 10-20 km.
Especially, under varying weather conditions the
flying crews often do decide in a short time which
project will be flown during the day. Further the
photogrammetric projects often cover larger areas,
so that multiple reference stations are necessary.
e the tropospheric errors which are still inherent in
airborne GPS data even if observation differencing
is used. Differencing can not eliminate the error
effects, because the tropospheric conditions
(temperature, pressure) in the aircraft and on the
ground reference station are usually clearly different.
* using AROF also flat turns have to be flown,
because longer continuous stretches of data are
required for a successful ambiguity resolution.
The major advantage of using AROF is, that in principle
no ground control points would be required if the correct
ambiguities could have been estimated. However, doing
a block adjustment entirely without ground control brings
up some other problems which need to be addressed:
e The datum transformation between the WGS 84 and
the mapping system needs to be known with cm
accuracy
e The geoid in the block area needs to be known with
cm accuracy, as the GPS heights are not
orthometric
e estimation of self calibration parameters in the block
adjustment is not possible without any ground
control points
* Quality control is very difficult (How can one find an
error in the camera focal length?)
The combined block adjustment (CBA) of GPS and
image coordinates is based on the idea, that GPS and
aerial triangulation can both determine the camera
perspective center coordinates. The concept is that the
GPS ambiguity resolution is done in the
blockadjustment, and although more unknowns have to
733
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
be estimated, the number of ground control points can
be reduced significantly. The GPS observations
strengthen the block so much, that normally a minimum
of 4 ground control points in the block corners are
sufficient for the adjustment. As the ambiguity resolution
is ususally done stripwise, losses of phase lock may
occur during the turns, hence no restrictions apply to the
normal flying behavior. Especially, steep turns may be
flown without paying attention to loosing the GPS
signals. GPS biases, like troposphere, ionosphere or
clock errors can also be modeled in the block
adjustment, so that the distance between the reference
station and the rover can be as much as 500 km. Also,
with the CBA there is no need for dual frequecy
obsevations, as the ambiguity resolution is supported wit
the image coordinate observations. The above
mentioned operational advantages have lead to the
conclusion that the combined block adjustment is still
the better choice for the use of GPS in photogrammetric
post-processing.
For the performance analysis of the GPS receivers, the :
cycle ambiguities have been determined with the
combined block adjustment method. Figures 4a and 4b
show the differences between the conventional block
adjustment and the GPS positions which have been
determined with the cycle ambiguities in the combined
block adjustment.
0.100 -
0.050 ff \ AAR AM a a.
0.000 A-¢
Differences [m]
0.050 {ASV A / "v.
-0.100 *
Image Nr. [-]
Figure 4a Differences (x,y,z) between Aerial
Triangulation and GPS Post-Processing Positions (Leica
SR 399)
0.100 =
0.050 A — Rf
0.000 À- /
fferences [m]
|
-0.050
D
-0.100
Image Nr. [-]
Figure 4b Differences (x,y,z) between Aerial
Triangulation and GPS Post-Processing Positions (Leica
9212-Aero)