Orientation and Free Network Theory of Satellite CCD Line-Scanner Imagery
Atsushi Okamoto, Kyoto University
Susumu Hattori, Fukuyama University
Hiroyuki Hasegawa, PASCO Corporation
Tetsu Ono, Kyoto University
COMMISSION III
KEY WORDS: Satellite CCD Line-Scanner Imagery, Affine Transformation, Iterative Approach,
Free Network Analysis
ABSTRACT
In 1992 Okamoto et al. proposed an orientation theory of satellite CCD line-scanner imagery based on affine
transformation. However, in order to employ this orientation approach, the central-perspective satellite CCD
line-scanner images are required to be transformed into affine ones and this image transformation cannot be
performed without errors due to height differences in the terrain. Therefore, the original orientation method could be
applied only for the case where the terrain was hilly and the image transformation errors were negligibly small. In
this paper, we will first present an iterative orientation method of satellite CCD line-scanner imagery where the
image transformation errors are corrected.by using the ground height information obtained in the previous iteration
step. Tests with many simulation models clarified that two iteration steps are quite enough to remove the image
transformation errors, even when height differences in the terrain exceed 4,000 meters. Next, the free network theory
of satellite CCD line-scanner imagery is constructed, because this problem is becoming of greater importance for the
precise analysis of satellite CCD line-scanner imagery having very high ground resolution (1m - 4m) which may
appear in a few years.
INTRODUCTION
The general orientation theory of satellite CCD
line-scanner imagery such as SPOT imagery can
rigorously be constructed based on projective
transformation. However, this theory itself is of little
practical use, because the attained accuracy is rather low
due to very high correlations among the orientation
parameters. In order to overcome this difficulty, an
orientation theory of one-dimensional affine images was
derived by Okamoto et al. in 1992, which can be
applied for the analysis of satellite CCD line-scanner
imagery by transforming the central-perspective images
into affine ones. However, this image transformation
cannot be performed without errors due to height
differences in the terrain. Therefore, a correction
technique of the image transformation errors is required
to be developed to employ the orientation method using
affine transformation for the analysis of satellite CCD
line-scanner imagery of mountainous terrain where the
image transformation errors are not negligibly small.
Also, this problem can be overcome by introducing an
iterative orientation approach using affine transfor-
mation, where the image transformation errors are
corrected by using the ground height information
obtained in the previous iteration step.
Very little has been written that provides a general and
rigorous approach to the free network analysis of
satellite CCD line-scanner imagery, though this
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
problem is becoming of greater importance for the
precise analysis of forthcoming satellite CCD
line-scanner imagery having very high ground
resolution (Im - 4m). However, if the orientation
problem of overlapped images is solved and the
geometrical characteristics of the one-to-one
correspondence relating the model and object spaces are
clarified, we can easily find linearly independent free
network vectors required for obtaining unbiased
solutions with minimum variance. In the general
orientation problem of one-dimensional affine images,
the absolute orientation can be expressed by a
two-dimensional affine transformation having six
independent coefficients and the equation of the
photographing plane in the object space coordinate
system. Thus, in the free network analysis of satellite
CCD line-scanner imagery, we need nine linearly
independent vectors which can be given by linearizing
the one-to-one correspondence between the model and
object spaces.
In this paper, the general orientation problem of
one-dimensional affine images is first briefly discussed.
Then, the transformation of central-perspective
line-scanner images into affine ones is described with
correction of the image transformation errors, and
simulation results of the proposed iterative orientation
approach are discussed. Regarding the free network
analysis of satellite CCD line-scanner imagery, the nine
linearly independent vectors are given and the practical
604
ORI
AGI
Th:
of al
(Ok:
>—N
C)
Fig
of a
Syste
relat
of tl
plan
Figu
three
sepa
detei
affin
syste
plan:
Equ:
orier
can
three
TRA
TIV.