n Sensors
radiometers
jometers are
|! has precise
iometers are
/C a base-to-
functions of
lectrically to
tween other
y overlapped
are obtained
the satellite.
-, nadir- and
other due to
> degrees are
let mode in
:n-numbered
1 with Joint
ich are based
ization, and
from ALOS
rates can be
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
Ititude (km)
=
690
685
-00.-75 -60 -45 30 -15 0 315-30 48 60 75 90
Latitude (deg.)
Figure 5. Relationships between latitude and nominal orbit
2.2 Expecting Disturbances Affected Geometric Accuracy
Figure 3 presents the frequency characteristics of PRISM’s
sampling, position and attitude determination sensors aboard
ALOS, and expected disturbances i.e., attitude fluctuations that
will directly affect the geometric accuracy of PRISM images
(modified from Iwata e/ al, 2002). Each line of PRISM will be
sampled in 0.37 msec intervals, which corresponds to 2.5
meters on the ground. Each scene will be composed of five
second observations, which corresponds to 35 km in the along
track direction. The expected attitude fluctuations will result
from thermal distortions of PRISM itself, large strictures such
as the solar paddle, PALSAR's antenna due to thermal environ-
ment changes on the orbit, and dynamics by driving the data
relay antenna (DRC) and pointing mirror of AVNIR-2.
To determine these fluctuations, the ALOS has orbit and
attitude sensors such as dual frequencies GPS Receiver (GPSR),
Star Tracker (STT), Angler Displacement Sensor (ADS), and
Inertial Reference Unit (IRU). Furthermore, the Precise
Pointing and geolocation Determination System (PPDS), which
is one of the ground processing systems, was developed to
determine highly accurate PRISM's pointing based on attitude
sensors (Iwata er al., 2002). It is still necessary to prepare
many GCPs worldwide to validate the orbit and attitude sensors
themselves, as well as to evaluate the geometric accuracy of
PRISM because its sampling frequency is the highest among
onboard sensors.
2.3 Calibration and Validation Plans
Figure 4 summarizes the working flows of calibration and
validation for PRISM and their evaluation items. The most
important item is the geometric calibration, which is required to
generate highly accurate DEMs by PRISM's stereo pair images.
First, we will evaluate the relative alignments between each
CCD for each radiometer, which were measured on the ground
before launching the ALOS in the pre-flight test. However,
they may be changed by launching, and also depend on the
thermal condition while in orbit. The attitude determination
accuracies, stabilities, and control accuracies will also be
evaluated for each sampling frequency in Figure 3. To evaluate
these items, we require many highly accurate GCPs worldwide.
Before the geometric calibration, we may require some
radiometric calibration ie. relative radiometric calibration,
because stripes are the main obstacles in image interpretation
and matching with GCP. Therefore, we have developed an
evaluation tool for relative radiometric correction using
available satellite images (Tadono et al., 2003).
The image quality evaluation is also important and uses
standard products such as level 1A, 1BI and 1B2 processed in
JAXA’s data processing center, EOC.
320
318 == Nadir-Backward
i Nadir-Forward
=310
t “5
—
NJ X
Co
10
308
306
304
tance (km
S
1
-90 -75 -60 -45 -30 -IS 0 15 30 45 60 75 90
Latitude (deg.)
Figure 6. Relationships between latitude and distance between
observed areas (Black: nadir-backward; gray: nadir-forward)
The processing software of DEM and ortho-rectified image
using PRISM stereo pair images is in development based on the
triplet image matching technique (Takaku er a/., 2003). This
software introduces semi-epipolar line estimation, aerial based
matching, automatic window size optimization, coarse to fine
techniques, etc. The generated DEM and ortho-rectified image
are defined as high-level products processed in Earth Observa-
tion Research and Application Center (EORC), JAXA. The
GCP will also be used to validate the high-level products.
3. STRATEGY OF PREPAIRING GCPS AND
EVALUATION ITEMS
3.1 Estimation of Simultaneous Observation Areas
Basically, a higher number of GCPs is preferred because the
observed images may be affected by cloud cover. To
effectively calibrate the PRISM geometrically, we should
consider strategies to prepare GCPs worldwide, which depends
on target evaluation items and their frequency characteristics.
Before preparing a GCP, we simulate the ALOS nominal orbit
to estimate the location of each of PRISM radiometer's
observation. We then describe the preparation of GCP with
evaluation targets of geometric calibration.
3.1.1 Orbit Simulation: To estimate the locations of
simultaneous observations in triple observing mode, we first
simulate the ALOS nominal orbit. Figure 5 describes the
relationship between latitude and satellite altitude on Geodetic
Reference System 1980 (GRS 80). which is an ellipsoid model
using the ALOS. The altitude of the ALOS is 691.65 km above
the equator, and the inclination angle is 98.16 degrees for the
nominal orbit elements. The nominal orbit radius of the ALOS
was calculated as a function of declination, and then subtracted
from the radius of the ellipsoid model at the same location. The
x-axis of Figure 5 was translated into the geodetic latitude.
Satellite altitude is changing 30 km north to south, which
produces an observation gap for each radiometer of the PRISM.
3.1.2 Calculation of Pointing Vector: Figure 6 illustrates
the relationship between the geodetic latitude of the nadir
looking radiometer and the distance between each radiometer
on the GRS 80 ellipsoid model. The black line indicates the
distance between nadir- and backward-looking radiometers.
The gray line represents the distance between nadir- and
forward-looking radiometers. These lines were calculated from
the orbit altitude in Figure 5 and the alignments of each
radiometer measured on the ground. Figure 6 depicts the
