COMPARATIVE TESTES OF MATHEMATICAL MODELS FOR ACCURACY POTENTIAL OF
POINT MEASUREMENTS IN IKONOS GEO IMAGE
e ONE.
S. Sadeghian ,M. J. Valadan Zoe} :
! Research Institute of National Cartographic Center (NCC), Tehran, IRAN, P.O.Box: 13185-1684, Sadeghian(@nec.neda.net.ir
Faculty of Geodesy and Geomatics Engineering, K.N. Toosi University of Technology, No. 1346, P.O.Box 19697, Tehran, IRAN,
Valadanzouj@KNTU.ac.ir
Commission III , WG I11/1
KEY WORDS: Modelling. High resolution, Accuracy, Satellite, Parameters, IKONOS, Image
ABSTRACT:
The number of high resolution satellite sensors for mapping applications is growing fast.
Successful exploitation of the high accuracy potential
of these systems depends on good mathematical models for the sensor modelling. High resolution image data increases the need for higher
accuracy data modelling
. The satellite orbit, position, attitude angles and interior orientation parameters have to be adjusted in the geometrical
model to achieve optimum accuracy with the use of minimum number of Ground Control Points (GCPs). But most of high resolution satellite
vendors do not intend to publish their sensor models and ephemeris
data. At present, however, the necessary camera model and precise
ephemeris data for Ikonos have not been published. There is consequently a need for a range of alternative practical approaches for extracting
accurate terrain information from Ikonos imagery.
Geopositioning accuracy of Ikonos panchromatic Geo image is investig
equations, polynomials, 3D affine, multiquadric functions and thin plate spline
Also orbital parameters of Ikonos satellite have been estimated and an orbit
ated in this paper. Rational functions, the DLT, SDLT, 2D projective
methods have been applied in tests with Ikonos Geo image.
al parameter model has been expanded that cover the Ikonos Geo
image. This Ikonos rigorous model uses basic information from the metadata and image file. This is followed by the results of various
geometric accuracy tests carried out with Ikonos Geo image using different parameters (9, 12, 15) and combination of control and check
points. The test area cover parts of West of Iran. For determining GCPs and independent check points (ICPs), 3D digital maps was used.
Taking into account the quality of GCPs, |
1. INTRODUCTION
After the successful launch and deployment of Ikonos-2 satellite
in September 24, 1999, EROS-Al in December 5, 2000,
QuickBird-2 in October 18, 2001, SPOTS in May 4, 2002 and
OrbView 3 in June 26, 2003, the era of commercial high
resolution earth observation satellites for digital mapping had
began. Successful exploitation of the high accuracy potential of
these systems depends on a comprehensive mathematical
modeling of the imaging sensor. An orbital parameter model can
be applied to stereo space imagery in order to determine exterior
orientation parameters. Unfortunately the ancillary data (position,
velocity vectors and angular rates) of the satellite platform have
not been provided with Ikonos imagery, therefore alternative ways
of camera modeling need to be employed. Recently, several 2D
and 3D approaches have been reported to tackle this issue
(Valadan and Sadeghian, 2003a, Sadeghian and Delavar, 2003;
Dowman and Tao, 2002; Fraser et al, 2002a, 2002b; Valadan et
al, 2002; Hanley and Fraser, 2001; Sadeghian et al, 2001a,
2001b; Tao and Hu, 2001, 2002). They do not require interior
orientation parameters or orbit ephemeris information. The image
to object space transformation solution is based only upon ground
control points (GCPs). This is an advantage for processing the
new high resolution satellite imagery (HRSI). In this paper the
possibility of using non-rigorous and rigorous sensor models for
2D ground point determination from Ikonos Geo image is
investigated.
or the Ikonos Geo, an optimal accuracy of 0.9 m was achieved using the orbital parameter model.
2. ORIENTATION MODELS
At this writing, Space Imaging has refused to release information
on the sensor model for Ikonos, as well as data on the precise in-
flight position and attitude of the imaging sensor. This means that
a large number of photogrammetric parameters are unknown and
not readily determinable from the imagery alone. The very long
focal length and narrow angle of view (0.93?) and swath (—11 km)
will likely make an orbital resection unstable, and even if many
GCPs and several images are used, an accurate solution might not
be possible. There is consequently a need for a range of
alternative, practical approaches for extracting accurate 2D and
3D terrain information from HRSI. In the following discussion,
Rational functions, the DLT, SDLT, 2D projective equations,
polynomials, 3D affine, multiquadric functions and thin plate
spline method are evaluated as potential approximate sensor
models to substitute for the rigorous physical sensor model. Also
in this paper the possibility of using a rigorous model (orbital
parameter modelling) for ground point determination were
explored and investigated (Valadan and Sadeghian, 2003b). The
orbital parameters and ephemeris data have been approximated
from meta data, image file and celestial mechanics. These
parameters then have been used in the orbital parameter modeling.
International Arc
Jet US
2.1 Rational Func
Under the model, :
of two polynomial
coordinates (X,Y,Z
x = PI(X,Y,Z)/P2C
m m2 m3
ijk
i=0 /=0 k=0
y = P3(X,Y,Z)/P4(
ml m2 m3
SS S. Ci
i=0 j=0 k=0
(1)
For example RFM
Rational with 14 te
Rational with 17 te
a,
x +
i+
by +
yt e
i+
Rational with 20 tei
An FE
Xm -—-—------
1 + Cy
b,+ b
y E Ea E
[oy
The rational func
ground coordinates
frame, pushbroom,
linear transformati
affine, 2D projecti
forms of the RFM,
2.1.1 Direct Linea:
Eleven linear ori
between 2D image
ast aX + a,