ı recorded
weakened
or bundle
onal cross
ui permet
nent dans
aramétres
Yt tous les
faisceaux
blocs.
Daten, die
ohne den
! operatio-
den, ohne
1. Background
Continual progress in GPS technology as well as in
receiver hardware and analysis software also leads
to increases in accuracy and cost saving in
photogrammetry. Improved control of survey
aircraft and aerial cameras contributes towards
reliability during the flight. The start of aerial strips
as well as end and side laps can be kept to more
accurately. Overlapping with neighboring strips can
also be controlled accurately for each photograph.
Conjugated photographs from neighboring strips
overlap completely lengthwise. It is considerably
easier to select tie points and the number of these
points is reduced.
For the subsequent bundle block adjustment, the
GPS and camera data recorded with relation to the
antennae positions or the projection centers is
processed for all exposure times. In the process a
few imponderables ensue: The GPS position is
determined in a fixed rhythm of e.g. half a second.
In this half second, the aerial survey craft covers a
distance of approx. 35 m. An interpolation over this
distance leads to some losses in accuracy.
During the GPS data analysis, due to the geometry
of the satellite configuration and the wave lengths of
the signals, phase ambiguities have to be resolved
in order to obtain correct results. For ground
stations this is done using a longer observation
time. In the aircraft, dependent on the number of
received satellites, the flight strip length, the
stability of the reception in the individual strips, the
inclination of the aircraft in the turning loops, which
could lead to loss of signals, and the constancy of
the weather conditions, it may be difficult or
impossible to resolve these phase ambiguities. In
individual cases this may result in incorrect
solutions. Such uncertainties and errors cause a
misalignment of the exposure positions and in
addition a falsification, which is essentially time-
dependent, of the coordinates. The altitude
components are particularly sensitive to those
errors.
As a result of the GPS analysis, the exposure
positions are obtained in geocentric coordinates in
the WGS84 datum. These are then converted to the
national coordinate system. However, there is not
always sufficient information for the datum
transformation.
In the subsequent bundle block adjustment, these
positions are considered as observations. In other
solution approaches which exist to date, in order to
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
recognize and rectify possible coordinate
falsifications, shift and drift parameters are
introduced into the block adjustment as additional
unknowns. 6 parameters are obtained for each flight
section if the satellite configuration remains the
same and the reception is constant. The position
function is replaced by the time. This is permitted in
the first approximation if the aircraft flies on a
straight course. Also the non-linearity of the
influences of the ambiguities can be modeled
without any disadvantages by a linear drift
formulation for flight strips which are not too long.
The number of additional unknowns which ‘weaken’
the block can quickly become very high.
The conditions required for the shift and drift
formulation of the unchanging satellite constellation
and the uninterrupted signal reception partly lead in
practice to considerable information loss. This is
because satellites which have only been observed
during a part of a flight section, or for which a cycle
slip has occurred, have to be removed from the
analysis. Due to this a weakening of the geometry
results which in turn causes a greater non-linearity
of the influence of the ambiguities as well as lower
accuracies.
To ensure the determinability of shift and drift
parameters, it is thus necessary to increase the
number of control points or to introduce cross strips
for block stabilization. The possible gain from
including GPS data is at least partly lost again.
The shift and drift parameters model, in addition to
the non resolved phase ambiguities, simultaneously
time errors from uncorrected datum transformations
and non modeled influences of a changing
troposphere. Even if individual shift and drift
parameters are not significant and thus could
actually be taken out of the adjustment, they often
have to be retained due to the aforementioned
reasons as they are physically justified.
A further problem is the correct weighting of the
GPS data in the bundle block adjustment. Due to
incorrect weighting formulations, accuracies and
reliabilities can easily be reproduced which have
nothing to do with the reality.
2. A new approach
A combination of the known software packages
GEONAP and BINGO-F considerably improves the
functional approach with the help of a rigorous
mathematical model. After the GPS data has been
395
ee
