1 2004
e that
gorous
> main
tation
often
and
ess is
d first
sation
>d and
ty are
ige of
nation
„GPS
1sors).
'borne
ymials
IS are
n be a
n data
lens
when
ficient
' class
? and
tellite
earity
e line:
ewise
itform
or and
id Tie
1 and
nts of
Zero,
>ctory
or all
order
duced
of the
or {”
f the
ed on
luded
ch is
ue to
Ld.
i5 P2)
) line
ed to
2840,
stereo
TER,
ternal
tions,
non-
io the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
first-order Taylor decomposition with respect to the unknown
parameters. The resulting system is solved with a least square
adjustment. As result the coefficients of the polynomials
modelling the external orientation, the self-calibration
parameters and the coordinates of the tie points are estimated.
Statistics on the system s
In case of satellite imagery, the available ephemeris (usually
sensor position and velocity at fixed intervals) are used to
generate the approximate values for the parameters modelling
the sensor external orientation (position and attitude). The
required geometric parameters (focal length(s), viewing angles,
number and size of CCD elements in each array) are usually
available from the imagery provider or from literature. The
reference frame used in the adjustment is the fixed Earth-
centered Cartesian system, also called ECR.
In case of airborne imagery, the GPS and INS observations are
included in the piecewise polynomial equations. The
polynomial coefficients model the shift and offset between the
GPS and INS local systems and the camera system (centred in
the lens perspective centre) and 1" and 2" order systematic
errors contained in the observations.
For the orientation of the pushbroom imagery preliminary tests
are made with these objectives:
e determination of best degree for piecewise polynomials and
best GCPs configuration, by solving the adjustment without
self-calibration, with quadratic functions modelling the
external orientation and varying the number of segments and
GCPs configuration;
e external orientation modelling with linear and quadratic
functions, using the best GCPs configuration and best
trajectory segments, without self-calibration;
e self-calibration with best external orientation modelling
configuration.
The choice of the unknown self-calibration parameters to
include in the modelling is based on the analysis of the cross-
correlation between the self-calibration parameters, the external
orientation parameters and the ground coordinates of the TPs.
Statistics on the adjustment performance and RMS values for
the GCPs and Check Points (CPs) are considered for the quality
assessment.
4. ORIENTATION OF SATELLITE IMAGES
The model has been applied for the orientation of satellite and
airborne images with different acquisition geometry (one-lens
and multi-lens optical systems, synchronous and asynchronous
acquisition) and ground resolution. As satellite applications
concern, in (Poli, 2003), (Giulio Tonolo et al., 2003) and (Poli
et al., 2004) the results obtained by the orientation of MOMS-
02/P, MISR, EROS-A1 and SPOT-S/HRS are presented, while
in (Poli, 2002) the tests carried on the Three Line Sensor (TLS),
carried on helicopter, are reported. In the following paragraphs
the latest results obtained from SPOT-S5/HRS and ASTER are
summarised.
4.4 SPOT-S/HRS
Within the HRS-SAP Initiative (Baudoin et al., 2004), a DEM
was generated from two stereo images acquired by the High
Resolution Stereoscopy (HRS) sensor carried on the newest
satellite of SPOT constellation. The sensor model was applied
in order to orient the stereopair and estimate the ground
coordinates of the CPs. The available ephemeris (sensor
position and velocity) were used to generate the approximate
values for the parameters modeling the sensor external
orientation (position and attitude) in fixed Earth-centred
geocentric Cartesian system. From the available 41 object
points, a group of them was used as GCPs and the remaining as
CPs. The best results in terms of RMSE in the CPs were
obtained by modelling the external orientation with two 2™
order polynomials and with self-calibration. The self-calibration
parameters that mostly influenced the model were &;, &», p» and
s, for both lenses. The other self-calibration parameters could
not be estimated due to the high correlation with the TP
coordinates and external orientation parameters. By changing
the number of GCPs and CPs, the RMSE were always less than
| pixel. For a more detailed description of the data and
processing, see (Poli et al., 2004).
4.2 ASTER
ASTER (Advanced Spaceborne Thermal Emission and
Reflection Radiometer) is a high-resolution, multispectral/
hyperspectral imaging instrument which is flying on Terra, a
satellite launched in December 1999 as part of NASA's Earth
Observing System (EOS). ASTER takes data in 14 spectral
bands within the Visible and Near Infrared (VNIR), the
Shortwave Infrared (SWIR) and the Thermal Infrared (TIR) at
ground resolution of 15m, 30m and 90m respectively.
The generation of DEMS is possible with the VNIR instrument,
that provided stereo images in along-trak direction. VNIR
consists of two independent telescopes operating in band 3
(0.76-0.87um), viewing nadir (channel 3N) and backward
(channel 3B, 27.6 off-nadir) with respect to the spacecraft
trajectory. The telescopes scan the ground in pushbroom mode
using arrays of CCDs with size 7um x 7um. The number of
CCD elements in each array is 4100 for channel 3N and 5000
for channel 3B (4100 are active). The two telescopes allow
simultaneous stereo imaging with a 64 sec time delay between
the scanning of the same ground target and a B/H of 0.6. Each
scene is 4100x4200 pixels large and cover an area of about
60km x 60km. Several types of ASTER data are available at
different processing levels. For our purposes, the level 1A is
used, because at this level the images are not geometrically
processed.
The scene used in this work covers the valley of Shaxr, in the
South-East part of China. The images were kindly provided by
the Institute for Spatial and Landscape Planning, ETH Zurich,
who is involved in a World Monuments Fund project for the
economic development of the Shaxi valley. The scenes were
acquired on 23" November 2000 in the morning.
For the orientation of the channels 3N and 3B from the ASTER
scene, six GCPs that have been used. The ground coordinates of
these points were available from in-situ GPS measurements or
measured in local maps. As the sensor external orientation
concerns, the ASTER scene metadata file contained the satellite
position and velocity in ECR (fixed Earth-centred Cartesian
coordinate system) every 400 image lines. These data were used
to calculate the satellite attitude at the observations times and
calculate the initial approximations for the polynomial
coefficients modelling the sensor external orientation. Due to
the limited number of object points, only the RMSE for the
GCPs have been calculated: 8.1m in X, 8.4m in Y and 10.4m in
Z. After the production of five pyramid images, interest points
were matched and found progressively in all pyramid levels
starting from the low-density features on the images with the
lowest resolution. After the process, 32Q,000 points were
successfully matched. The failed matches were mostly in
correspondence of areas covered by clouds, due to the cloud
movement between the nadir and backward images acquisition.
