A Practical Detection of Non-Cooperative
Satellite Based on Ellipse Fitting
Yang Liu, Zongwu Xie, Bin Wang
*
and Hong Liu Zheng Jing, Cong Huang and Shi Bao
State Key Laboratory of Robotics and System Institute of Manned Space System Engineering
Harbin Institute of Technology Beijing, 10009, P.R.China
Harbin, 150001 Heilongjiang, P.R.China huangcongbest@sina.com
liuyanghit@hit.edu.cn
Abstract – As some of the valuable satellites can maintain
working after on-orbit servicing, it's meaningful to study on the
autonomous rendezvous. Most of these satellites are non-
cooperative, more specifically, they have no target location
markers or rendezvous sensors. This paper extracts the edge of
the adapter ring as natures for recognition, measurement or even
mechanical docking. A practical detection method of non-
cooperative satellite based on ellipse fitting is proposed in this
paper. By detecting the contours of the input image, we can
estimate the ellipse parameters at the curve level. Linear
segmentation method is used to extract straight-line segments
from edge contours. After that, we construct a graph model to
segment each line contour into elliptical arcs of a possible ellipse.
Finally, ellipse detection based on least square algorithm is used
to get the expression of the quadric curve in the pixel coordinate
system of the image. The corresponding algorithm is
experimented on the image in all directions with complex
backgrounds, it can provide a real-time result with high accuracy
compared with conventional methodologies.
Index Terms – ellipse detection; non-cooperative target; on-
orbit servicing; adapter ring; real-time
I.
I
NTRODUCTION
As of January 1, 2016, the Union of Concerned Scientists
(UCS) lists 1381 active satellites in orbit. 493 out of 1381
satellites carry out their mission in GEO (geostationary orbit).
Taking the “inactive” objects (including abandoned satellites,
rocket bodies and space debris) into consideration, only about
32% objects in GEO are under controlled [1].
Some GEO satellites will go to wrong orbits due to launch
failure, which will result in large economic cost and other bad
impacts. In fact, most of these satellites can be rescued by
auxiliary orbits maintain, or simple repairs [2]. Therefore,
many countries and organizations have studied on the on-orbit
servicing technologies in GEO including repairing, upgrading,
refueling and re-orbiting spacecraft on-orbit [3].
Canadian development has led space robotics activities on
the Space Shuttle Columbia on STS-2 in November 1981 with
the Shuttle Remote Manipulator System (SRMS). Japanese
space robotics development has also been substantial, having
flown a free-flying LEO spacecraft with two robotic arms to
the ISS. The DLR (German Aerospace Center) Institute of
Robotics conducted its first space robotic experiment flew on
a Spacelab mission on STS-55 in 1993. The United States
flew an on-orbit servicing demonstration called Orbital
Express (OE) to LEO in 2007. The program was
experimented with the leadership of DARPA with
participation by NASA. Recently, China has also launched a
lunar mission, the Yutu Moon Rover (also called the Jade
Rabbit) that includes a small robot arm to position an alpha-
proton x-ray spectrometer.
As the space shuttle or other manned spacecraft can’t reach
GEO, servicing satellites in GEO needs ability to rendezvous
and dock by an unmanned spacecraft with a high degree of
autonomy. Non-cooperative satellite usually equips no target
location marker or rendezvous sensor, which results in the
incompleteness and imprecision of the observation
information. Hence, pose measurement of target satellite
becomes a challenging task in a GEO servicing mission [4].
Over the last decade, vision-based navigation systems have
been widely used to settle the problem of relative pose
determination, due to their low cost, mass, and power
requirements, compared to active sensor-based techniques.
Besides, it has a better accuracy when the target and service
satellite are close relative [5].
Generally speaking, the satellites already existing in GEO
are not designed to be serviced, i.e. no artificial marks used
for measurement are mounted on them [6] [7]. Therefore, we
have to rely on the presence of natural features on the target
satellite. Among common features of the mechanical design of
conventional satellites, the adapter ring (also called interface
ring) is a good candidate to be grasped [8]. It is a high-
strength torus with a radius of either 1194 mm or 1666 mm
[9], which is used to connect the satellite to the launch
vehicle.
Therefore, ellipse detection of the adapter ring is a key
point in the accomplishment of non-cooperative satellite
rendezvous. Based on the results of the ellipse detection, we
can determine the relative pose of the target. Based on their
architecture, ellipse detection algorithms can be categorized
into three groups [10].
First group is based on Hough Transform. Image is first
transformed into parametric space, then we can estimate the
parameters of the ellipse by finding peaks in the parametric
space. These methods usually have a large amount of
computational costs and a waste of storage space. The
accuracy depends on the edge detection of the ellipse.
Second group mainly uses least square fitting method to
minimize the sum of algebraic or geometric errors. These
methods put much emphasis on how to accurately fit an
1541
978-1-5090-2396-7/16/$31.00 ©2016 IEEE
Proceedings of 2016 IEEE
International Conference on Mechatronics and Automation
August 7 - 10, Harbin, China