
4.9 System for Measurement of Position and Attitude 241
Depending on the desired applications, the main focus has to be on higher posi-
tioning performance or alternatively higher quality in attitude determination. This
ultimately defines the quality requirements for GPS/IMU exterior orientations (see
Fig. 4.9-1), as well as the approaches taken for GPS/IMU processing (i.e. real-time,
post-processing, GPS processing [Table 2.10-1)].
For digital camera systems, the maximum tolerable error in object space is lim-
ited by the sensor’s ground sampling distance (GSD). Assuming that the smallest
object to be identified in the images has to be at least the size of one object pixel,
the required accuracy of object point determination also has to be in the range of
one object pixel or better. From this the following equation can be found:
h
g
= c ·m
b
= c ·
GSD
pix
=
c
pix
·GSD = k ·GSD. (4.9-1)
The resulting flying height above ground h
g
is a function of camera f ocal length
c and sensor pixel size pix. The k factor is obtained from the quotient of these. Its
reciprocal value
1
k
=
pix
c
is known as the instantaneous field of view (IFOV) of the
sensor. Equation (4.9-1) also shows that in the case of digital imaging the GSD plays
a similar role to image scale in data acquisition from the former analogue imagery.
Integrated GPS/IMU systems used for the direct orientation of airborne sensors
are typically used in the following circumstances. The update information from GPS
is available throughout the whole mission flight. If any satellite blockages and signal
loss of lock are present, they appear mainly during flight turns. Sequences without
any GPS update information are relatively short. Within the remaining parts of the
trajectory, enough GPS data is available between two signal loss of lock events to
resolve the integer phase ambiguities reliably. This is mandatory in the case where
differential carrier phase processing is required and applied. Such considerations
are of importance in the case of decentralized GPS/IMU data processing. Here raw
GPS measurements (pseudo-range, doppler and phase observations) are not used
for update, but already processed GPS position and velocity data (so-called pseudo-
observations) are fed into the filter. This requires a minimum of four satellites to
provide update information. Alternatively, if the GPS/IMU processing is performed
within a centralized filtering approach, raw GPS observations are used as update
information. Thus updates are possible even within periods where less than four
satellites are available. Nevertheless, GPS/IMU systems used in airborne appli-
cations for direct sensor orientation are mostly based on the decentralized filter,
owing to the higher flexibility of decentralized filters if additional components are
integrated, i.e. updates from other sensors. This is different to centralized filters,
where major parts of the algorithm have to be redesigned for changes in system
configuration.
If such integrated GPS/IMU systems for measurement of position and attitude
are used in dynamic environments, a certain bandwidth and sampling rate have to
be achieved to describe the dynamics of the sensor’s movement sufficiently. Since
the high frequency parts of the dynamics are measured by the inertial sensors, the
IMU specifications are of major concern. Within Fig. 4.9-3 two different spectra