PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
November 2016
893
Geometric Calibration of Ziyuan-3
Three-Line Cameras Using Ground Control Lines
Jinshan Cao, Xiuxiao Yuan, Yi Fang, and Jianya Gong
Abstract
Owing to the large biases of laboratory-calibrated imaging
parameters, the initial direct georeferencing accuracy of Ziyu-
an-3 (
ZY-3
) three-line camera (
TLC
) images can only reach the
kilometer level. In this paper, a line-based geometric calibration
model is established by using the line-coplanarity constraint,
and a feasible geometric calibration approach for the
ZY-3 TLC
s
using ground control lines (
GCLs
) is presented. Experimental re-
sults show that the calibration accuracy achieved by using
GCLs
reached about 1.0 pixel, which was almost the same as that
achieved by using ground control points (
GCPs
). After substitut-
ing
GCLs
for
GCPs
in the geometric calibration, the georeferencing
accuracy of the
ZY-3 TLC
images without ground controls was
significantly improved from the kilometer level to better than
11 m in planimetry and 8 m in height. A better georeferencing
accuracy of about 3.0 m in planimetry and 2.5 m in height was
achieved after the block adjustment with four
GCPs
.
Introduction
Precise georeferencing is a prerequisite for many applications
of high-resolution satellite (
HRS
) imagery. Georeferencing
accuracy determines the geometric quality of the products
derived from
HRS
imagery, such as digital elevation models
(
DEMs
) and digital orthophoto maps (
DOMs
). In order to realize
the precise georeferencing of
HRS
imagery, we should first
establish a suitable sensor orientation model to relate image
space and object space. Existing sensor models can mainly
be classified into two categories: physical sensor models and
empirical sensor models (Habib
et al.
, 2007; Poli, 2005 and
2007; Toutin, 2004). Physical sensor models are based on the
collinearity equations that can explicitly reflect the physical
characteristics of
HRS
sensors. Empirical sensor models, such
as the rational function model (
RFM
), are based on polynomial
functions. They are often used when the parameters of the
imaging system or physical sensor models are unavailable.
When we perform the direct georeferencing of
HRS
imagery
using physical sensor models, we need to know accurate im-
aging parameters of
HRS
sensors, including interior orientation
parameters (
IOPs
), exterior orientation parameters (
EOPs
), and
mounting parameters. The
IOPs
describe the internal character-
istics of
HRS
sensors, including focal length, principal point,
lens distortions, etc. The
EOPs
describe the satellite positions
and attitudes, and they are usually measured by global posi-
tioning system (
GPS
) receivers, star trackers, and gyroscopes.
During the satellite’s launch and in-orbit operation, some of
the imaging parameters of
HRS
sensors, such as camera mount-
ing angles (rotation angles from the camera coordinate system
to the satellite-body coordinate system; Wang
et al.
, 2014),
often change more or less. These changes may be caused by
high acceleration, the variation of spatial environments, and
some changes in the geometric arrangement of the optical
instruments. The biases of the
IOPs
and the mounting param-
eters will consequently lead to systematic errors in the direct
georeferencing. Therefore, geometric calibration is an indis-
pensable step in exploring the full geometric performance
of
HRS
sensors. It aims to obtain accurate
IOPs
and mounting
parameters. With the help of geometric calibration, a better
and more satisfactory direct georeferencing accuracy can be
guaranteed during the whole lifetime of
HRS
sensors.
For the majority of currently available
HRS
sensors, such as
Ikonos,
SPOT-5
, OrbView-3, IRS-P6, and
ALOS/PRISM
, many spe-
cific geometric calibration approaches have been developed,
as referred to in Gachet (2004), Grodecki and Dial (2002),
Kocaman and Gruen (2008), Mulawa (2004), Radhadevi and
Solanki (2008), and Radhadevi
et al.
(2011). Although these
approaches may differ from each other because of the different
structural characteristics of different satellite sensors, their
objectives stay the same: to improve image georeferencing
accuracy. For example, the direct georeferencing accuracy of
IRS-P6 images was improved from about 2000 to 250 m after
calibration of the individual sensor alignment, the inter-cam-
era alignment, and the focal plane (Radhadevi and Solanki,
2008). The georeferencing of
ALOS/PRISM
images after self-cal-
ibration with 30 additional parameters achieved almost the
same accuracy (expressed in pixels) as that achieved by
SPOT-
5
, Ikonos, and QuickBird images (Kocaman and Gruen, 2008).
Similarly, several approaches have been developed to
geometrically calibrate Ziyuan-3 (
ZY-3
) three-line cameras
(
TLCs
). Using a large number of ground control points (
GCPs
)
extracted from the reference
DOMs
and
DEMs
, Wang
et al.
(2014)
calibrated the look angles of the charge-coupled device (
CCD
)
detectors and camera mounting angles. Note that the
CCD
-de-
tector look angles here refer to the two viewing angles of a
projection ray in the camera coordinate system. Likewise, us-
ing the reference
DOMs
and
DEMs
as ground controls, G. Zhang
et al.
(2014) calibrated the offset matrix and the errors in the
principal point displacement, the focal length, the size of the
CCD
detector, and the rotation angle of the
CCD
array. Y. Zhang
et al.
(2014) calibrated the
CCD
misalignment and the misalign-
ment between the
ZY-3
TLCs
and the satellite. Chen
et al.
(2015)
calculated the
IOPs
and
EOPs
of the
ZY-3
TLC
images using the
self-calibration bundle adjustment. Although the above ap-
proaches differ from each other, all of them can significantly
improve the georeferencing accuracy of the
ZY-3
TLC
images.
In general, when we perform the geometric calibration of
HRS
sensors,
GCPs
are employed as ground controls. The
GCPs
can be highly accurate fixed landmarks in the calibration field
or can be extracted from the reference
DOMs
and
DEMs
cover-
ing the calibration field. For the former, the number of
GCPs
is
limited, and the construction and maintenance of fixed
Jinshan Cao is with the Collaborative Innovation Center of
Geospatial Technology, 129 Luoyu Road, Wuhan, China,
430079 (
).
Xiuxiao Yuan, Yi Fang, and Jianya Gong are with the School
of Remote Sensing and Information Engineering, Wuhan Uni-
versity, 129 Luoyu Road, Wuhan, China, 430079.
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 11, November 2016, pp. 893–902.
0099-1112/16/893–902
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.11.893
