International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
2. SATELLITE-BASED STEREO ANALYSIS 
The satellite-based stereo analysis includes several processing 
steps which are illustrated in Figure 1. After the data description 
in Section 2.1, the sensor model for image georeferencing will 
be presented in Section 2.2. The subsequent processing steps 
are explained in Section 2.3. 
  
  
  
  
  
  
e Sensor model with external 
GEOREFERENCING orientation estimation and self- 
calibration 
à À 
PREPROCESSING e dencepyramid 
e Wallis filter 
À À 
  
e Fórstner or Harris operator 
FEATURE SELECTION 
S e Thinning with cloud mask 
X^ 
[ MATCHING | e Hierarchical LSM 
  
  
  
  
  
X 
QUALITY CONTROL 
X 
| PRELIMINARY CTH, | e Prata and Turner formula 
X 
[CTW CORRECTION | + CTW from Meteosat-6/-7 
e Absolute and relative tests on 
LSM matching statistics 
  
  
  
  
  
  
  
  
X 
[ FINAL CIH: | 
  
Figure 1. Schematic overview of stereo-photogrammetric 
processing of the satellite-based images to derive 
CTH and CTW. 
2.1 Data 
2.1.1 ATSR2 / AATSR: The Along Track Scanning 
Radiometer (ATSR2) instrument is part of the ERS-2 satellite 
system which was launched in April 1995. The successor 
sensor, AATSR, is part of Envisat, which was launched in 
Spring 2002. ERS-2 and Envisat are in a near-circular, sun- 
synchronous orbit at a mean height of 780 km, an inclination of 
98.5? and a sub-satellite velocity of 6.7 km/s. The repeat cycle 
of ATSR2/AATSR is approximately 3 days. 
The ATSR2/AATSR sensor first views the surface along the 
direction of the orbit track at an incidence angle of 55? as it flies 
toward the scene. Then, some 120 s later, ATSR2 records a 
second observation of the scene at an angle close to the nadir. 
The ATSR2 field of view is comprised of two 500 km-wide 
curved swaths with 555 pixels across the nadir swath and 371 
pixels across the forward swath. The pixel size is 1 km x | km 
at the center of the nadir scan and 1.5 km x 2 km at the center of 
the forward scan. The sensor records in seven spectral channels, 
i.e. 0.55 um, 0.67 um, 0.87 pm, 1.6 pm, 3.7 pm, 10.8 pm and 
12.0 pm. All channels have a radiometric resolution of 10-bit. 
Our CTH retrieval is based on the rectified data products, GBT 
for ATSR2 and ATS. TOA IP for AATSR. The geolocation of 
these rectified products is achieved by mapping the acquired 
pixels onto a 512 x 512 grid with 1 km pixel size whose axes 
are the satellite ground-track and great circles orthogonal to the 
ground-track. 
2.1.2 | MISR: The Multi-angle Imaging SpectroRadiometer 
(MISR) is currently the only operational satellite that acquires 
images from nine different viewing angles. MISR was launched 
on board the EOS AM-1 Terra spacecraft in December 1999. 
The orbit is sun-synchronous at a mean height of 705 km with 
an inclination of 98.5? and an equatorial crossing time of about 
10:30 local solar time. The repeat cycle is 16 days. The MISR 
instrument consists of nine pushbroom cameras at different 
viewing angles: -70.5? (named DA). -60.0* (CA), -45.6? (BA), - 
26.1? (AA), 0.0? (AN), 26.1? (AF), 45.6? (BF), 60.0? (CF), and 
70.5? (DF). The time delay between adjacent camera views is 
45-60 seconds, which results in a total delay between the DA 
and DF images of about 7 minutes. The four MISR spectral 
bands are centered at 446 nm (blue), 558 nm (green), 672 nm 
(red) and 866 nm (NIR). The red-band data from all nine 
cameras and all spectral bands of the nadir camera are saved in 
high-resolution with a pixel size of 275 m x 275 m. The data of 
the blue, green and NIR bands of the remaining eight non-nadir 
cameras are stored in low-resolution with a pixel size of 1.1 km 
x 1.1 km. The operational data products from MISR are 
described in (Lewicki et al., 1999). The two products used for 
this study are the LIB1 radiance and the LIB2 ellipsoid- 
projected radiance data. 
The LIBI product is radiometrically but not geometrically 
corrected, while the L1B2 ellipsoid-projected radiance product 
is referenced to the surface of the WGS84 ellipsoid with no 
terrain elevation included. The MISR georectified product 
spatial horizontal accuracy requirements are driven by the needs 
of the geophysical parameter retrieval algorithms. The goal of 
operational MISR data processing is to achieve an uncertainty 
better than + 140 m for both the absolute geolocation of the 
nadir camera and the co-registration between all nine cameras 
(Jovanovic et al., 2002). The latest evaluation results of the 
L1B2 geolocation accuracy as shown in (Jovanovic et al, 2004) 
are approaching prelaunch requirements, with along- and cross- 
track errors far below | pixel for all cameras (except DA). 
The operational L2TC top-of-atmosphere/ cloud product, which 
contains the operationally derived cloud parameters, like stereo 
CTH, east-west (EW) and north-south (NS) cloud motion 
components, as well as many additional parameters from the 
stereo retrieval (Diner et al., 2001), can be used as comparison 
data for validation (Seiz, 2003). 
2.2 Sensor Modeling 
The aim of rigorous sensor models is to establish a relationship 
between image and ground reference systems according to the 
sensor geometry of acquisition. In particular, different 
approaches have been proposed for the georeferencing of 
pushbroom sensors carried on aircraft (Gruen et al., 2002) and 
satellite (Poli, 2003). A flexible sensor model that can be 
applied to a wide class of linear CCD array sensors has been 
developed in our group and already applied to different linear 
scanners carried on satellite and aircraft (Poli, 2003). The model 
is based on the photogrammetric collinearity equations, that are 
extended in order to include the external orientation modeling 
with 2" order piecewise polynomials and a self-calibration for 
the correction of lens distortions and CCD lines rotations in the 
focal plane. 
1156 
crit 
mil 
cro 
diff