IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 14, NO. 10, OCTOBER 2017 1765
Ship Classification in Moderate-Resolution SAR
Image by Naive Geometric Features-Combined
Multiple Kernel Learning
Haitao Lang, Member, IEEE,andSiwenWu
Abstract— Compared with the high-resolution synthetic aper-
ture radar (SAR) image, a moderate-resolution SAR image can
offer wider swath, which is more suitable for maritime ship
surveillance. Taking into account the amount of information in
a moderate-resolution SAR image and the stability of feature
extraction, we propose naive geometric features (NGFs) for
ship classification. In contrast to the strictly defined geometric
features (SGFs), the extraction of NGFs is very simpler and
efficient. And more importantly, the NGFs are enough to reveal
the essential difference between different types of ships for
classification. To fuse various NGFs with different physical prop-
erties and discriminability, the multiple kernel learning (MKL) is
utilized to learn the combination weights, rather than assigning
the same weight to all features as usually applied by the
traditional support vector machines (SVMs). The comprehensive
experiments validate that: 1) the performance of the proposed
NGF-combined MKL outperforms that of NGF-combined SVM
by 3.4% and is very close to that obtained by SGF-combined
MKL and 2) in terms of classifying ships in a moderate-resolution
SAR image, NGFs are more feasible than scattering features.
Index Terms— Multiple kernel learning (MKL), naive geomet-
ric features (NGFs), remote sensing, ship classification, synthetic
aperture radar (SAR).
I. INTRODUCTION
S
PACE-BORNE synthetic aperture radar (SAR) can
provide wide-area ob servation capability in nearly all
meteorological phenomena and atmospheric conditions and
has been extensively used in maritime surveillance [1].
Knowledge about the positions and types of ship targets may
serve for a wide range of applications, such as maritime traffic
management, fisheries control, and border surveillance [2] .
Ship de tection in SAR images has been intensively studied
and has been developed as a crucial component of many oper-
ational ocean monitoring systems, such as SUMO, SIMONS,
and so on [3]. In recent y ears, the advent of the new generation
of satellite missions
1
makes it possible to further identifying
Manuscript received June 12, 2017; revised July 20, 2017; accepted
July 25, 2017. Date of publication September 1, 2017; date of current version
September 25, 2017. This work was supported in part by the National Natural
Science Foundation of China under Grant 61471024, the National Marine
Technology Program for Public Welfare under Grant 201505002-1, and the
Higher Education and High-Quality and World-Class Uni versities under Grant
PY201619. (Corresponding author: Haitao Lang.)
The authors are with the Department of Physics and Electronics, Bei-
jing Unive rsity of Chemical Technology, Beijing 100029, China (e-mail:
haitaolang@hotmail.com).
Color versions of one or more of the figures in this letter are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LGRS.2017.2734889
1
Readers who are interested in these missions are referred to a special issue
of IEEE Geosci. Remote Sens. Mag., vol. 2, no. 2, 2014.
the type of ships beyond normally detecting/locating the
targets.
As mentioned in [4], to achieve high-performance ship clas-
sification, the discriminating feature extraction from the ship
signature in an SAR image and the optimized classifier design
are two critical issues that should be carefully addressed.
Recently, some attempts to classify ships in a single polar-
ization high-resolution SAR image
2
have been carried out.
Margarit and Tabasco [5] combined the scattering features,
[i.e., the mean radar cross section (RCS) value retrieved for
the bow, middle, and stern sections of the ship signature]
and the geometric features (the length and width of the ship)
to categorize ships based on a fuzzy logic decision rule;
Xing et al. [6] proposed to combine different geometric
features (i.e., length , length-to-width ratio, shape complexity,
centroid, and the covariance coefficient) and the scattering
features (local RCS density of several parts of a ship) to rep-
resent the ship and utilized the sparse repr esentation method
to classify ships; and Wang et al. [7] presented a hierarchical
classifier to classify ships, also based on the geometric and
scattering features.
Since an SAR image swath is inversely proportional to
image resolution, a high-resolution SAR image is often
obtained in the cost of narrow swath. Typically, a high-
resolution SAR image at the resolution of about 1 m often
corresponds to a swath of 5 km. Such narrow swath is unsuit-
able for maritime surveillance, which objectively requires
a wide-area observation capability. For maritime surveil-
lance applications, today’s space-borne SAR offers swaths of
100–450 km at image resolution o f 10–30 m. Thus, in the
view of applications, it is necessary to explore the feasibility of
classifying (or preclassifying) ships in a moderate-resolution
SAR image.
As aforementioned, geometric features and scattering fea-
tures have been widely used for ship classification in a high-
resolution SAR image. However, when considering classifying
ships in a moderate-resolution SAR image, the geometric fea-
tures are mo re advantageous than the latter. As shown in Fig. 1,
these subfigures show the same ship (MMSI:355587000)
3
acquired in different imaging conditions: resolution, polariza-
tion, incidence, and so on. Comparing Fig. 1(a)–(d), it shows
2
The study of this letter focuses on ship classification in a single polarization
SAR image. Readers who are interested in classifying ships with polarimetric
features can refer to the related studies.
3
The detailed information of this ship can be obtained by visiting the website
http://www.marinetraffic.com.
1545-598X © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.