d", b I son M nudis E sins
S
= INS
2 g E^ ues V nio
5 > | MASTER FILTER
= Strapdown »
S INS - (8r^,5v',e" d^ b^
5 o' Mechanization
= ib
5 » A
2 position
2 velocity
: UPDATE attitude
r INS» Vans
phase, Doppler correct GPS calibration
seudorange measure-
p g N ments FILTER
(6r', öv‘)
Cycle Slip
Detection
phase, Doppler
<
pseudorange
E Vers
Figure 1: Block Diagram of Sensors Integration
of the problem can be supplied by a precise attitude/positioning
system. Furthermore, direct exterior orientation allows the
georeferencing of remotely sensed data in near real time (Schwarz et
al. 1993).
Position and attitude accuracies needed are application dependent.
The horizontal and vertical accuracy on the ground is mainly
determined by the accuracy of the cartographic reproduction process
and the contourline interval of the maps. The contourline interval is
mainly dependent of the scale of the maps and the slope of the terrain
and can range from 1 m to several tenths of meters. Assuming that
contourlines are not allowed to intersect, the vertical accuracy has to
be at least half of the contourline interval (Schwidefsky/Ackermann
1976). The cartographic reproduction quality is determined by the
map production facilities and therefore might range from 0.1-
0.25 mm. These requirements can be much more stringent if remote
sensing is applied for cadastral point determination or engineering
tasks. In this case position accuracies better than 10 cm are required
in object space. Taking these positioning accuracies into account the
Map Scale / Horizontal Vertical Attitude
Application Accuracy Accuracy Accuracy
[m] [m] [107 deg]
1:50 000 10 8 35
1:25000 5 4 30
1:5000 1 0.75 15
Cadastral
Point <0.1 «0.1 5
Determination
Table 1: Required Positioning and Attitude Accuracies
(RMS Values)
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
required accuracies for the attitudes can be calculated. The attitude
parameters are mainly dependent on the flying height abouve ground
and the focal lenght of the sensor. Assuming a wide-angle aerial
camera attitude accuracies as given in Table 1 are necessary for the
orientation process.
Two accuracy classes will be considered in the following. For
cadastral or precise engineering projects, accuracies have to be at the
sub-decimetre level for position and at the level of five milli-degrees
or better for attitude. For mapping applications at the scale of
1:10000 and smaller and for many resource mapping applications
with multi-spectral scanners, accuracies at the level of one metre or
less for position and of a few tenths of a degree for attitude are
sufficient. In this paper the potential of integrated GPS/INS systems
is evaluated for the direct determination of the exterior orientation
parameters for these levels of accuracy.
In the described test, the feasibility of directly determined exterior
orientation parameters is evaluated in two steps. First, the in-flight
orientation accuracy and position of the integrated GPS/INS is
assessed by comparing it to orientation parameters independently
determined by inverse photogrammetry, i.e. by using a large number
of accurate ground control points to determine position and attitude
parameters at aircraft level from the image measurements. Second,
coordinates of pre-surveyed check points on the ground are
determined by georeferencing independent models whose exterior
orientation has been derived from the integrated GPS/INS system.
2. THE SENSOR INTEGRATION
In order to obtain the best positioning/attitude performance, the INS
data are integrated with GPS double differential measurements in a
decentralized Kalman filter configuration (Figure 1). The GPS filter
is independent of the INS filter and its output is used to update the
INS error states. The double difference pseudorange, carrier phase
and phase rate observations form the measurement vector in the
GPS filter. Its output (i.e. position and velocity) is taken as a set of
126
pseudo-mea
filter. The no
GPS filter co
filter are fed
the strapdow
position, vel:
to check the
carrier phase
. lock in the (
including 1
correlated se
timevarying
modelled b:
includes pc
augmented t
equation de
INS mechan
and Wei (1¢
Since the G]
from the
displaceme
GPS/INS 1
perspective
the translatic
techniques
Similarly, a
between I^
account to
perspective
more comp
physically
implement
frame cam
imagery (C
calibration
between the
can be achi
rigid platfor
the aerial c
The measu
at the expos
the high ra
derived pc
displaceme
between the
N
TN AC
Thus, after
with sens
Equation 2
can be rep:
3. TE
A well-de
Germany