processing
d real time
natic deri-
contours.
mote Sen-
gh fidelity
eparation
eneration,
ity digital
1 Remote
1tegration
) Geogra-
h. Photo-
. 116-121.
‚ordnung,
«tion und
otogram-
jn, Reihe
cess data
[nt. Arch.
art 4, pp.
or Digital
qualitati-
ammetri-
:he Kom-
Effect of planimetiric correction with a few GCPs on terrain
height estimation with stereo pair
Kohei Arai, S.Mukai and Y. Terayama
Department of Information Science, Saga University
1 Honjo, Saga-city, Saga 840 Japan
+81-952-24-6010 (FAX)
arai@is. saga-u. ac. jp(E-mail)
+81-952-24-5191 Ext. 2567,
Abstract
Effect of planimetric correction
with a few GCPs on terrain height
estimation with stereo pair has been
confirmed. It is well known terrain
height estimation accuracy depends on
geometric fidelity, in particular,
planimetric distortion. If a few GCPs
with almost same terrain height are
extracted from stereo pair, planimetric
distortion can be corrected results in
an improvement of terrain height
estimation accuracy.
1. Introduction
In Digital Elevation Model (DEM)
estimation with stereo pair, estimation
accuracy depends on stereo matching
accuracy. It is also obvious that stereo
matching accuracy depends on planimetric
fidelity of stereo pair. In order to
confirm it, | an attempt has been
conducted with system corrected SPOT/HRV
of stereo pair of imagery data.
In this study, it is assumed that a
few GCPs(Ground Control Points) are
available with not so accurate location
accuracy and with almost same height. A
method for improvement of stereo
matching accuracy with such a few GCPs
is proposed. Effects of the proposed
method is evaluated in terms of DEM
estimation accuracy and planimetric
accuracy.
2. The proposed method
In general, DEM is estimated with
stereo pair of chip images including the
point of interest through stereo
matching. Before that, a geometric
correction, generally, is applied to the
raw images such as system correction.
The proposed method is to apply a simple
planimetric correction such as Affine
transformation with a few GCPs of not so
accurate location accuracy and of almost
same sea level before the stereo
matching.
System corrected images have
residual geometric distortion results in
low planimetric fidelity. This implies
poor stereo matching accuracy. If the
planimetric fidelity is improved, then
stereo matching accuracy will be
improved results in increasing DEM
857
estimation accuracy. In the planimetric
correction, it can not be expected that
a number of GCPs with high location
accuracy is existing. Because if such a
number of GCPs are available, it is
expected that a high accurate large
scale map is available so that it is
also expected that a high accurate DEM
can be estimated from such a map.
Assuming such a large scale map is not
available and also assuming that there
is only a few GCPs with not so accurate
location accuracy extracted from a small
scale map, a method for improvement of
planimetric fidelity is proposed.
In order to improve a planimetric
fidelity, the following first order
Affine transformation is assumed.
[L' |=|01|+|cosQ —sinQ||1 -sinR|
IP” | |Op| |sinQ cosQ||0 cosR|
l1 0 | Me iil (1)
[0 Mil IP]
where (L,P) and (L’,P’) are line and
pixel coordinate of GCP in the
planimetrically distorted system
corrected image and planimetrically
corrected image, Ol, Op are offsets in
line and pixel directions, Me, Mi are
magnification and aspect ratios and Q, R
denote skew and rotation, respectively.
Ignoring the sencond order terms from
equation (1), (L’,P’) is expressed as
follows.
|L’|=|01| + Me |1 —Mi(Q+R) | IL| (2)
IP" | lOpl . IQ Mi | IPI
=]A11 A12| ILI + IB1] (3)
|A21 A22| |P| |B2]
The coefficients of the above equation
are determined with a few GCPs so that
if a few GCPs with almost same height
are available, then planimetric
distortion will be removed through
Affine transformation with the
determined coefficients.
The aim of the proposed method is
to improve DEM estimation accuracy
through planimetric correction.
3. Experimental consideration with
SPOT/HRV stereo pair
