Satellite-Imagery Geometric Accuracy
Improvement Based on Direct Correction of
Dominant Coefficients
Xinming Tang, Changru Liu, Ping Zhou, Ning Cao, Fengxiang Li, and Xia Wang
Abstract
An important and difficult point in the application of satellite
imagery is refining the positioning model and improving the
geometric accuracy. In this study, we focus on improvement
in geometric accuracy and develop a new rational function
model (
RFM
) refinement method. First, we derive the conver-
sion relationship between the rigorous sensor model and the
RFM
, based on which we illustrate the approximate meaning
of the zero-order and first-order terms of the rational poly-
nomial coefficients (
RPCs
). Second, the correlation problem
between
RPCs
and the influence of individual
RPCs
on geo-
metric positioning accuracy are analyzed and verified. The
dominant coefficients that determine geolocation are then
identified. Finally, a new
RFM
refinement method based on
direct correction of the dominant coefficients is proposed and
validated. The experiments, conducted with
ZY3-02
satel-
lite imagery, indicate that the proposed method can effec-
tively improve the geometric accuracy of satellite images.
Introduction
In recent decades, with the continuous development of satel-
lite technology, numerous high-resolution optical satellites
have been put into use, providing an effective means for rapid
and large-scale collection of geospatial information. In particu-
lar, with the successful launch of satellites such as the United
States’
Ikonos
(Fraser, Hanley and Yamakawa 2002; Ager 2003;
Lutes 2004),
QuickBird
(Noguchi
et al.
2004), and WorldView
series (Aguilar, Saldaña and Aguilar 201
tem (Madani 1999), Japan’s ALOS (Osaw
Ziyuan-3 (X. Tang
et al.
2013; Xu
et al.
2
series (Zheng and Zhang 2016), remote sensing satellite im-
ages have been increasingly used in digital model generation
(Toutin 2004; Poon
et al.
2005), topographic mapping (Di, Ma
and Li 2003; Holland, Boyd and Marshall 2006), environmen-
tal protection (Stumpf, Malet and Delacourt 2017), and more,
producing significant economic and social benefits. However,
the imaging model of the satellite image is the key technology
in processing and application of remote sensing images. How
to refine the imaging model to improve geometric accuracy is
an important and difficult research point.
Nowadays, the rational function model (
RFM
) is the main-
stream postprocessing model of most satellites, owing to its
advantages of sensor independence, computational conve-
nience, and universality, and it is the first choice to replace
the rigorous sensor model (
RSM
; Tao and Hu 2001; Grodecki
and Dial 2003; Fraser, Dial and Grodecki 2006). Usually, the
RFM
provided by satellite-imagery vendors is calculated from
an
RSM
with direct least-squares solutions. However, due to
the effects of incomplete calibration and measurement er-
rors of satellite orbit and attitude, inherent errors are always
introduced into the
RFM
, which affect the geometric position-
ing accuracy. To eliminate these errors, various compensation
methods have been proposed.
Regenerating the rational polynomial coefficients (
RPCs
)
with control data is the simplest solution (Tao and Hu 2001;
Hu and Tao 2002; Long, Jiao and He 2014). This scheme is
simple in theory and effective when the
RSM
is excessively
complicated to develop or is not available. However, this
method is obviously terrain dependent, and it is highly
dependent on the actual terrain relief and the number and
distribution of ground control points (
GCPs
), which are dif-
ficult constraints in practical applications. When the accuracy
requirement is not stringent, this method can be used for pho-
togrammetry rectification and remote sensing applications.
Correction parameters attached to the
RFM
are the most
commonly used bias-compensation method at present. These
bias-compensation models are defined in image space or
object space and usually modeled as shift, shift and drift,
affine transformation, and second-order polynomial models
(Fraser and Hanley 2005). Aguilar
et al.
(2013) and Fraser
and Hanley (2003) have demonstrated that the shift model is
effective for
Ikonos
, WorldView, and GeoEye satellites. As for
QuickBird
images, Noguchi
et al.
(2004) have stated that the
shift-and-drift model is warranted if the highest possible ac-
, Liu, and Weng (2010) and J. Wang, Di,
ed shift, shift-and-drift, affine transfor-
rder polynomial models using
Quick-
Bird
and
Ikonos
imagery and obtained meter-level accuracy
with a number of good-quality
GCPs
. Experiments by Pan
et
al.
(2013), T. Wang
et al.
(2014), and Liu
et al.
(2013) con-
firmed that the affine transformation model is effective for the
Ziyuan-3 (
ZY3
) and
TH1
satellites. Similar research has been
reported for other types of satellite images with this biase-
compensation method (Jiang
et al.
2015; Jeong
et al.
2016;
Topan
et al.
2016; M. Wang
et al.
2017).
Although these bias-compensation methods can improve
geometric positioning accuracy, they are implemented with
additional parameters and the original
RPCs
remain un-
changed; thus, the geometric positioning often depends on
these correction parameters. In addition, for compatibility
with different photogrammetric systems it is often necessary
to generate new
RPCs
using these correction parameters (Fraser
et al.
2006; Tong
et al.
2010; Rupnik
et al.
, 2016). Although
Xinming Tang, Changru Liu, Ping Zhou, Ning Cao, and Xia
Wang are with the Land Satellite Remote Sensing Application
Center, Ministry of Natural Resource of the People’s Republic
of China, Beijing 100048, China (
).
Fengxiang Li is with the Shandong Provincial Institute of
Land Surveying and Mapping, Jinan 250102, China.
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 4, April 2020, pp. 215–224.
0099-1112/20/215–224
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.4.215
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
April 2020
215
