stanbul 2004
ased on the
widely used
software and
ension (Geo)
packages.
o-referencing
| (geographic
ind may also
tion and map
in Petroleum
to the World
j) Language).
HTML and it
d associated
read directly
ects "in XSI
isforms and
EM supplied
binary (.bil)
jeotiff with a
er left X and
or the user.
ocument file
een supplied
| points have
ir geographic
ordinates are
M (WGS 84,
is said to be
'E
> for Indian
oped at SAC
d for DEM
as originally
1C/1D data.
nodel of the
perspective
in the same
coordinates
shed through
f coordinate
ed to process
996).
version 9
as also used
MAP format
of the image
geometric
fication, and
netric model
etric model
for Remote
is model is
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part Bl. Istanbul 2004
based on principles relating to orbitography, photogrammetry,
geodesy and cartography.
3.3 Erdas Imagine: Erdas /magine Software version 8.6
developed by Erdas Inc, USA has been used for the
interpolation, re-projection and evaluation of the DEMs
generated. Even though Orhtobase Pro module of Erdas
Imagine can be used to generate DEM from different satellite
stereo pairs, as Version 8.6 of this software available at SAC
does not support the DIMAP format, Erdas Imagine was not
used for DEM generation.
4. METHODOLOGY ADOPTED
4.4 Stereo Data Preprocessing: Inputs required for the
processing of stereo data are the following:
e Ephemeris information i.e. position (X,y,Z), and velocity
(x. dot, y dot, z dot) of satellite at regular intervals of time
e Spacecraft attitude (in roll, pitch and yaw directions)
e Sensor pointing angles
e Mission/ Payload parameters
e Ground and Image coordinates of Ground Control Points
(GCPs).
These required information are given in a file called
metadata.dim in XML format. This file is read and different
input parameters have been extracted and stored in various files
in required formats for processing with Saphire. These inputs
were directly read by the Geomatica software. For Montmirail,
as there was no separate GCP control available, the individual
scene corners and the centre coordinates have been used as
GCPs in Saphire. In Geomatica, in addition to these points three
more control points were added using manually identified
relative control points after an affine transformation performed
using the given corner coordinates.
4.2 Processing Steps Involved for Saphire: Process of DEM
generation is divided into the following steps:
1. GCP Identification; 2. Space resection; 3. Conjugate point
identification; 4. Space intersection; 5. DEM interpolation.
4.2.1 GCP Identification: For data set 1 (Montmirail) no GCPs
are available. So four corner coordinates are taken as reference
points and used as GCPs. For data set 2 (Melbourne) 30 GCPs
are available. These GCPs are all concentrated in a small
portion (13 km x 11 km) of the image (120 km x 60 km).
Ground coordinates of GCPs are given in carth-centered earth-
fixed geocentric coordinates system and correspond to UTM
zone 55 Southern hemisphere coordinates. These points are
identified on the image to get corresponding image coordinates.
4.2.2 Space Resection: Space resection is basically modeling
and updating the position and orientation of the satellite, using a
few GCPs and proper geometric model relating the ground
coordinates of the GCPs and their corresponding image
coordinates identified on the image. Space resection procedure
used here in this exercise is based on the principle of
collinearity condition given below:
of X A EH X S
-xzk[M]|* Y,- Y,
us 2 Z s
where, (f. -x, -y) are the image coordinates, (X, Ya Zu) are the
object coordinates of the image point and (Xs Ys Z3 are the
coordinates of the perspective center for a particular imaging
line and & is the scale factor. M is the rotation (orientation)
matrix from one co-ordinate system to the other, which is a
function of exterior orientation parameters i.e look angle,
attitude and orbital parameters. Out of these only attitudes (Roll
(Qa), pitch (7) and yaw (y) ) are considered for modeling. Roll
(a), pitch (B) and yaw (y) are assumed to be linearly varying
with time for a short duration of time. Therefore these may be
written as:
Q— at at,
B- py Bit.
Y-n»*t.
Collinearity equation is nonlinear with respect to the attitude.
Therefore this equation is linearised by Taylor series expansion.
Now correction in attitude and their rates, that is Aq, Aa; Af)
AB, Ay. Ay, are estimated by solving the linearised collinearity
equations. Then these corrections are added to the initial
approximations and revised values of parameters are computed
as follows.
ay 7a, 4 Aa,
Bo =Bo +40. Bi - BI AB,
ncn am MS HAN
This procedure is applied iteratively till the collinearity
condition is satisfied. Thus correction/ updating of attitude is
performed iteratively.
1 0
Q; 0 t4Aa;
The difference in the convention of directions of attitude axes
between IRS-1C/ID and SPOT-5 and also the SPOT-5 images
having along track stereo as against IRS-1C/1D having across
track stereo, necessary changes were incorporated in order to
satisfy the collinearity condition.
4.2.3 Conjugate Points Identification: Conjugate points are
identified on a stereo pair of images in automatic mode by
hierarchical matching technique. Accuracy of DEM normally
depends on the accuracy of the conjugate point pairs identified
on stereo images. Following are the basic steps of hierarchical
matching: 1. Interest point identification; 2. Local mapping; &
3. Digital correlation. The image is subjected to a low pass filter
of varying size to get images of various scales (levels of
hierarchy) like 16, 8, 4, 2 or 1. At each level interest points are
identified on the reference image and correlation of these points
are done in the other image. Match points of a particular level
are used to establish local correspondence between the stereo
pair of images at the next level. Then by digital correlation,
exact location of interest points in the other image are
computed. This procedure is repeated from the highest level (16
times sub-sampled imagery) up to the last level (which is full
resolution image). Match points obtained in the last level are the
conjugate points used for the DEM generation. These conjugate
points are on the image space and hence they are not regularly
spaced with respect to the ground.
4.2.4 Space Intersection: To locate a position of an image
point in the three dimensional object space, the object must be
imaged from two different exposure stations. Procedure of
computing a ground coordinates from a stereo pair of images is
called space intersection. Collinearity condition equations are
the basis for computing the 3D ground coordinates of conjugate
points.
f X aros
- x =i PX, Es
- y zo mm
This equation can be rewritten as follows,
