International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
2. OVERVIEW OF THE VHR SATELLITE IMAGE
DATA AND GEOMETRIC PROCESSING
2.1 Basic characteristics of the sensors
Table 1 gives an overview of operationally active VHR satellite
systems used in the JRC 2003 campaıgn.
Features / sensors | Ikonos QuickBird EROS A
Launch Date .|24 Sept.1999 | I8 Oct.2001 | 5 Dec.2000
Satellite Altitude | 681 km 450 km 480 km
Resolution [m] |0.82 pan 0.61 pan 1.8 pan
(GSD - in nadir) |3.28 ms 2.44 ms
Image Swath
: : 11.3 km 16.5 km 13.5 km
(in nadir)
Revisit Time —
e 6 days 8 days 7 days
(40? lat... 15?
off-nadir)
Dynamic Range | 11-bits /pixel | 11-bits /pixel | 11-bits /pixel
Table 1. Basic technical parameters of VHR satellites in 2003.
2.2 Distortions in image geometry
An image is a collection of single lines registered continuously
by pushbroom line scanner. In the direction of the linear array,
a perspective projection can be defined and along the
perpendicular (to linear array) direction a parallel projection is
present. The exterior orientation for each line in the image is
different, but with regard to the level of regularity and stability
of satellite orbit, the change can be considered as a function of
time. The following geometric distortions are related to the
image formation process (Toutin et al, 2002):
* distortions caused by the platform and mainly related to the
variation of the elliptic movement around the Earth (position,
velocity, and attitude),
distortions due to the imaging sensor (the calibration
parameters, such as the focal length and the instantaneous
field of view; the panoramic distortion in combination with
the oblique viewing system, the Earth curvature and the
topographic relief changes the ground pixel sampling along
the column);
distortions due to the Earth (the rotation generates lateral
displacements in the column direction between image lines
depending of the latitude; the curvature creates variation in
the image pixel spacing; the topographic relief generates
parallax in the scanning azimuth)
In addition to the above mentioned distortions,
deformations arise during the georeferencing process, i.e. the
approximation of the geoid by reference ellipsoid and the
projection of reference ellipsoid on the tangent plane.
Most of distortions (with the help of system related data) is
corrected at the ground receiving station, but others are the
subject of further processing, often done by the end user. More
detailed considerations about VHR image distortions and
correction methods as well are included for example in: Toutin
2003, Toutin et al, 2002, Grodecki & Dial 2001,
2.3 Imagery product levels for orthorectification
Different product levels for a given satellite system are
available on the market. The type of product is defined by
radiometric and geometric pre-processing levels, with some
influence on pricing. Some products (not tested here) are
already orthorectified with high accuracy but they have a
corresponding higher price, as well as usually a requirement to
supply ancillary data to the image provider. It is generally
considered that the most appropriate processing levels for
creating accurate 3D geometric correction are:
Ikonos Geo orthokit: geometrically corrected and rectified
to a specified ellipsoid and map projection, supplied with Image
Geometry Model (camera information, RPC), enabling the
complete and accurate sensor geometry at the time of the image
collection. The pre-processing removes image distortions
introduced by the collection geometry and re-samples the
imagery to a uniform ground sample distance and specified map
projection. GEO has 15 m (CE 90%) standard horizontal
accuracy, excluding effect of terrain displacement, (SI, 2004).
QuickBird ortho ready standard: radiometrically
corrected, sensor corrected, geometrically corrected, and
mapped to a cartographic projection. No topographic
corrections applied. Provided with RPCoefficients enabling
orthocorrection. Standard Imagery products have a positional
accuracy of 23- meter (CE 90%), excluding any topographic
displacement (Eurimage, 2004). Ground reference is based on
refined satellite attitude and ephemeris information without
requiring the use of GCPs.
EROS 1A: radiometric system correction - calibrated and
gain adjusted to correct for known radiance response
characteristics of the camera sensor system, no geometric
system correction. No RPC data is available from the image
provider.
2.4 3D geometric correction methods
The geometric correction process for VHR satellite images,
unlike high resolution images, is somewhat sensitive and needs
more accurate ancillary data. This is due to the sensor (image)
parameters, acquisition conditions, and potentially achievable
target planimetric accuracy. The 2D polynomial based approach
— often sufficient for geometric correction of high resolution
images — is no longer useable for VHR images if the
commensurate accuracy of final product is intended; the
significant (in relation to image GSD) level of distortion -
especially relief displacement — demands 3D geometric
correction (orthorectification) methods. Such methods can be
divided basically into two categories:
parametric: rigorous (physical, deterministic) sensor
modelling with mathematical modelling of viewing geometry
physical components (platform, imaging sensor, earth, map).
Such models are complicated due to the information released
(or not) by image suppliers, although approaches exist to
overcome this problem e.g.: Toutin's model for VHR satellite
images, available in PCI Geomatica software.
non-parametric: the Rational Functions mathematical model
(RF), that builds a correlation between the pixels and their
ground locations (continuous mapping between image and
object space) based on ratio (separately for row & column) of
two cubic polynomial functions. Polynomial coefficients (the
rational polynomial coefficients - RPC) are derived using
physical sensor/camera model (at the ground station) and are
distributed by image vendor with certain processing level
products.
The chosen approach to orthorectification depends
frequently on available ancillary data, and the possibilities ol
the software accessible to the user. Both constraints have
implications on the choice of image type and its processing
level. The following options concerning the geometric model
for image correction can be considered in practise (Table 2):
.
1020
Interna
pein
"Model
Rigoro
satellit
LL ————
Ration:
calcula
(IGM /
Refine
model
deliver
Table
Th
effecti
intend
Se
2001)
GCPs.
geopo:
betwee
geome
Of
careful
accura
(during
orthorc
Earth
feature
or cha
the DE
31 T
With 1
positic
extrem
Systen
image:
area :
import
and a
CAP
manag
Th
paralle
intend.
with tl
* estal
diffe
RMS
* dete;
relat
data
* conf
relia
