Extrapolated Georeferencing of High-Resolution
Satellite Imagery Based on the Strip Constraint
Jinshan Cao, Xiuxiao Yuan, Jianya Gong, and Miaozhong Xu
Abstract
Ground control points (
GCPs
) are necessary in order to achieve
precise georeferencing of high-resolution satellite (
HRS
)
imagery. However, measuring
GCPs
is costly, laborious, and
time consuming. In some remote areas, we cannot even obtain
well-defined
GCPs
. In this study, a strip constraint model is
established. Based on the bias-compensated rational function
model and the strip constraint model, a feasible extrapolated
georeferencing approach for
HRS
imagery is presented. The
presented approach remains effective even when the interme-
diate images in the strip are unavailable. Experimental results
of the two ZiYuan-3 (
ZY-3
) nadir datasets show that the direct
georeferencing accuracy of the
ZY-3
nadir images reaches only
9 to 12 pixels. With four
GCPs
in the first image, the georefer-
encing accuracy of the other images in the strip is improved
to better than 2 pixels through extrapolated georeferencing.
Introduction
In order to improve the direct georeferencing accuracy of
high-resolution satellite (
HRS
) imagery, currently available sat-
ellites are often equipped with global positioning system (
GPS
)
receivers, star trackers, and gyros.
GPS
receivers are used to
determine satellite positions. Star trackers and gyros are used
to determine satellite attitudes. With the help of satellite posi-
tions and attitudes, we can usually achieve a good direct geo-
referencing accuracy of
HRS
imagery after in-orbit geometric
calibration (Cao
et al
., 2015; de Lussy
et al
., 2012; Grodecki
and Dial, 2002). However, owing to the systematic errors in
interior orientation parameters (
IOPs
) and satellite positions
and attitudes, the best georeferencing accuracy often cannot
be achieved without ground control points (
GCPs
) (Poli, 2007;
Teo, 2011; Weser
et al
., 2008; Yuan and Yu, 2008).
Nowadays, most of
HRS
sensors are linear array sensors.
An extensive strip image can be collected by a satellite sensor
through the push-broom mode. The strip image is subse-
quently divided into many standard images and provided to
users. If we perform indirect georeferencing of these standard
images separately, sufficient and evenly distributed
GCPs
are
required in each image to achieve the best georeferencing ac-
curacy. Measuring
GCPs
is costly, laborious, and time consum-
ing. It may be very difficult and even impossible for us to
acquire well-defined
GCPs
in each standard image, especially
in remote areas, cloud-covered areas, and forest areas. Con-
sidering these difficulties, previous studies have investigated
precise georeferencing of
HRS
strip images in order to drasti-
cally reduce the required number of
GCPs
.
Chen
et al
. (2006) performed indirect georeferencing of two
FORMOSAT
-2 strips of 15 and 12 standard images; they achieved
an accuracy of about 2.5 pixels with 33 and 25
GCPs
in the two
strips. Fraser
et al
. (2006) tested block adjustment of six Ikonos
stereo strips, and an accuracy of about 1.0 pixel was achieved
with only two
GCPs
. Rottensteiner
et al
. (2009) performed strip
adjustment of an
ALOS/PRISM
strip of 21 standard images. An
accuracy of better than 1.0 pixel was achieved based on a strip
constraint, whereas the required number of
GCPs
was reduced
from 44 to four. Fraser and Ravanbakhsh (2011) further evalu-
ated the georeferencing accuracy of a longer
ALOS/PRISM
strip
of 55 images; they achieved one-pixel level accuracy with only
four
GCPs
at the beginning and the end of the strip. Zhang
et al
.
(2009) performed block adjustment of a
SPOT5
stereo strip, and
an accuracy of better than 1.0 pixel was achieved with six
GCPs
.
Similarly, Cheng
et al
. (2010) tested block adjustment of four
long stereo strips from
SPOT5
, and an accuracy of better than 1.0
pixel was also achieved with eight
GCPs
. Nagasubramanian
et
al
. (2007) reported indirect georeferencing of two IRS-P6 LISS-4
long strips; they achieved an accuracy of about 1.5 pixels with
four
GCPs
. Zhang
et al
. (2014 and 2015) and Pan
et al
. (2016)
tested block adjustment of two ZiYuan-3 (
ZY-3
) triplet stereo
strips, and an accuracy of better than 1.0 pixel was achieved
with sparse
GCPs
at the beginning and the end of the strips.
Apparently, previous studies have obtained impressive results
in precise georeferencing of different
HRS
strips. With sparse
GCPs
at the beginning and the end of the strips, a satisfactory georefer-
encing accuracy can often be achieved. In this case, independent
check points (
ICPs
) are actually distributed within the ground
coverage of
GCPs
when we evaluate the georeferencing accuracy
using
ICPs
, which can be seen in Fraser and Ravanbakhsh (2011),
Pan
et al
. (2016), Rottensteiner
et al
. (2009), and Zhang
et al
.
(2014 and 2015). That is to say, previous studies mainly focused
on interpolated georeferencing of
HRS
strips. However, a
HRS
strip
image often covers a long ground area. Sometimes, we can only
acquire
GCPs
at one end of the strip, especially when the other
end is in remote areas. In this situation, extrapolated georeferenc-
ing perhaps can provide a better solution to improve the georef-
erencing accuracy of those standard image without
GCPs
.
In this study, based on the bias-compensated rational
function model (
RFM
) and a strip constraint model, we pres-
ent a feasible extrapolated georeferencing approach for
HRS
imagery. With the help of sparse
GCPs
in the first image of the
strip, the georeferencing accuracy of the other images can be
effectively improved. The remainder of this paper is orga-
nized as follows. The flowchart of the presented approach is
introduced, and the major procedures are described in detail.
Then, two
ZY-3
nadir datasets are used to analyze the feasibil-
ity of the presented approach. Finally, the conclusions of this
paper are presented.
Jinshan Cao is with the Collaborative Innovation Center
of Geospatial Technology, 129 Luoyu Road, Wuhan,
China, 430079.
Xiuxiao Yuan and Jianya Gong are with the School of Remote
Sensing and Information Engineering, Wuhan University, 129
Luoyu Road, Wuhan, China, 430079 (
).
Miaozhong Xu is with the State Key Laboratory of Information
Engineering in Surveying, Mapping and Remote Sensing,
Wuhan University, 129 Luoyu Road, Wuhan, China, 430079.
Photogrammetric Engineering & Remote Sensing
Vol. 83, No. 7, July 2017, pp. 493–499.
0099-1112/17/493–499
© 2017 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.83.7.493
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
July 2017
493
