more efficient real-time realisation, but extended with one more
parameter for improved precision. This model assumes a simplified
orbital, which could be approximated to a plane circle with adequate
accuracy during the imaging period of a scene. The orbit radii
determined from the ephemeris are fitted to a third-degree polynomial
to represent the deviations of the orbit from this circular shape. The
constant term of the polynomial is allowed to change in the adjustment,
but the linear, quadratic and third degree terms are kept fixed. The
orbital parameters used are thus reduced to four:
Inclination (I), Right Ascension (£2), Time at Ascending node (t = t 0),
Orbital radius at Ascending node (ro).
But the central travel angle is derived by v = (t - t?) 217P, where P is the
orbit period which for SPOT is 101.46 hours. A ‘leader’ (*.lea) file is
delivered with SPOT imagery. This leader file contains a predicted
ephemeris and measured angular velocities. Initial orbital parameters
could be derived from a block adjustment of ephemeris data. The
angular velocities measured about every 82 lines (125 ps) are assumed
to be of high, but not sufficient, accuracy to model attitude variations
of the imaging platform. Relative attitude angles in kappa, phi and
omega are computed by integration, but are interpolated into an 80-
lines spacing (76 values) to ensure a fast look-up access. This is crucial
to the real-time implementation for the Leica mapping terminal (LMT).
With these relative attitude angles, only the constant offsets in kappa,
phi, and omega are left to be determined. A linear rate of change for
phi is added for a total of 8 parameters. Linear rates for kappa and
omega were found to be insignificant, and the same was found for
quadratic rates of change for the three attitude parameters. Simulations
showed that the Westin model does not sufficiently model SPOT
without at least a linear phi parameter. The orientation model is usually
set-up for a SPOT-panchromatic image. A SPOT-XS image set-up
would only require minor refinements for its doubled size of imaging
elements and reduction by half in number of pixels across- and along-
track. If unchanged, results will still be similar.
3.0 SPOT FUNCTIONAL MODEL
The adjustment is done in the earth centred inertial geocentric co-
ordinate system (ECI), but transformations are required between these
other systems:
° The earth centred, earth fixed geocentric system (ECEF).
° A local geodetic system with a known relationship to
geographical co-ordinates.
° À sensor co-ordinate system as described in section 3.21.
° À local orbital system which coincides with the attitude reference
system when all three attitude values are zero.
Control information would normally be available in a local geodetic
coordinate system. These must be tranformed to geographical
coordinates, then the ECEF, and finally to the ECI system before they
could be used in this model.
The link between a ground control point and its image co-ordinate is:
Xg = Xs + ARiRbRsxP
where
° xP = image co-ordinates vector.
° Xg = ground co-ordinates vector in ECI
° Xs = satellite position vector in ECI
° Au. -= scale factor
° Ri = Rotation from the orbital reference system to the ECI
° Rb = Rotation between the attitude reference system and the
orbital system. The interpolated attitude changes from the
measurement system are used here. Only the values of kappa, phi,
and omega offsets at the beginning of the scene are calculated in
the adjustment.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
e Rs = Rotation between the sensor and the attitude reference
system; the primary rotation in omega is the mirror inclination; it
would also take care of the CCD-sensor off-sets, if any. SPOT is
no longer publishing angular CCD off-sets, usually derived from
calibration.
Rs is optional because the system can accomodate for the absence of
this rotation in the model. Accordingly, the mirror inclination angle can
be used as an initial value for omega in Rb instead of a zero value.
The reverse equation is:
xb E 1/A .R .(Xg- X0)
where R z(Rs . Rb . Ri)
In matrix form, this becomes:
0 Ta ea os Xm Xo
mne ,. 4
v, EAE me 5n e yo y
zc 3 LES 133 Z5 Zo
e
yp is the y-image co-ordinate. and c is the focal length of the SPOT
camera. The SPOT header file provides a scene-center-time from which
imaging times for any lines could be calculated using the CCD
integration time which is 1.5004 us for SPOT. Radial data from the
ephemeris data is fitted to a third degree polynomial with respect to
time. Ephemeris data is also used to calculate the time at the acending
node, enabling the dynamic calculation of the travel angle at any image
point. The radius at any image point could be used to generate the
camera position co-ordinates by the inverse transformation with the
orthogonal matrix of keplerian rotation angles (Ri); see below. Of the
keplerian rotations parameters, Right ascension (Q) and Inclintion (I)
remain fairly constant; only the travel angle (v) is changing rapidly
with time.
AX, - XQ) + 1, (X,- X0) + 1, (X, - X0)
F = 0 = —C® :
nX,-X,) +1, (X,-X,) t n, (X, - X,)
F. (Xe) + D, (X, - X4) + La (X, - X,)
= ) = —C® E - =
s " : Bi (Xe - XQ) + Ba (X 7 Xe) + I3 (X. X)
The adjustment could correct five observations y-image co-ordinate,
time of imaging GCP (t, equivalent to x-pixel coordinate), Longitude
(9). Latitude(A) and height (h).:
[yp, t, 9, A, h].
The 8 parameters of orientation to be corrected are:
[lo, Qo, to, ro, wo, po, Ko, pto].
pto is the linear component of the phi rotational parameter. A Taylor's
series expansion of the collinearity equations are done but only the first
order terms are taken.
8k on 3r Om oF 3
Bo Op 81. 3
OF, OF, OF, OF, OF
>
2t
Il
«
=
Sy, à op X Oh
[9r 8k OF 3k am OR or àk |
p . |9 9 à, à e, 50, Ok d, |
| |8E Bk OE Bk GF KH OE
|ó 40 & à im 5 Sq $i
The conc
where v
paramete
misclosu
equation
Westin |
(mainly
with ‘fic
A
where \
paramete
observat
from ex]
effect the
Q is the
apriori w
V =
v can al
a few ite
3.1
The nee
required
been do
develope
superim[
The soft
making
resource:
Window:
permane
was stop
exactly v
towards
3.11
normally
in an an
marked i
co-ordin:
of the S
centre ha
3.12
geograpt
possible
for the ir
transforn
and inert