Ship Detection and Recognition based
on Multi-physical Fields
LI Yongqiang, LI Tie, YAN Wei, CUI Dong
Science and Technology on Electromechanical Dynamic
Control Laboratory
Xi’an, China
ZHAO Qi
, DING Mingli
School of Electrical Engineering and Automation
Harbin Institute of Technology
Harbin, China
Abstract—The detection and identification of the kinds of
ships, i.e., warship or merchant ship, is of great interest for
military use. Ships are usually detected and recognized based on
ship physical fields, and the commonly used ship physical fields
include sound field, magnetic field, hydraulic pressure field,
electric field, gravity field, etc, which all contain plenty of
discriminative information. However, the existing ship
identification methods are usually based on single ship physical
field, which will limit the model performance. In this work, we
proposed a ship recognition method based on combination of
multi-physical fields, i.e., sound field and magnetic field. To the
best of the authors’ knowledge, this is the first work to recognize
ships based on multi-physical fields. We fuse features extracted
from multi-physical fields by Principal Component Analysis
(PCA), which is then fed to Support Vector Machine (SVM) to
realize the recognition of ship. Plenty of experiments
demonstrate the effectiveness of the proposed method.
Keywords-ship physical fields; sound field; magnetic field; ship
recognition; SVM
I. INTRODUCTION
Because of great potential application value for marine
military, the recognition of ships has attracted great attention in
the past two decades. Ships are usually recognized from ship
physical fields. The ship physical fields include sound field,
magnetic field, hydraulic pressure field, electric field and
, which contain plenty of discriminative
information for ship detection and recognition. Usually only
one single physical field is used to realize the recognition of
warships and merchant ships, which will significantly limit
the recognition performance. In this work, we propose a new
ship recognition method based on multi-physical fields, i.e.,
sound field and magnetic field. To the best of the authors’
knowledge, this is the first work to recognize the kinds of
ships based on multi-physical fields.
Collecting real ship physical fields signal data is difficult
since the data acquisition of the ship physical field signal
based on the sea trial is quite expensive and time consuming.
Modeling and simulation of ship physical fields can overcome
the shortcomings of sea trial since simulation does not require
lots of financial and human resources. Hence, in this work we
are going to simulate and model ship physical fields, i.e.,
sound field and magnetic field, and then test the proposed
method based on the simulation data.
In this work, we first extract features from sound field and
magnetic field separately, then we extract features from both
sound field and magnetic field, which is further fused based
on Principle Component Analysis (PCA). We employ SVM as
the classifier to recognize the kind of ships. As a new machine
learning method, SVM can solve the small sample, nonlinear,
high dimension and local minimum points, and is widely used
in pattern recognition. The obtained feature vectors are
divided into training set and testing set. The training set is
used to train the classifier and the testing set is used to verify
the accuracy of the classifier. Radial basis function (RBF) was
chosen as the kernel function prediction model, and through
cross-validation method, the best parameters are determined.
Experimental results demonstrate that using ship sound field
and magnetic field together significantly improved the
recognition accuracy when
d to using sound field or
magnetic field separately.
II. M
ODEL AND SIMULATION OF SHIP PHYSICAL FIELDS
Since we do not have the real ship physical fields data, we
are going to model and simulate the ship sound filed and
magnetic filed in this section.
A. Modeling the sound field of ship
For modeling and simulating the ship-radiated noise, Ross
developed a regression function [1]. Researchers performed
spectral analysis to reflect the characteristic parameters of a
ship. Envelope power spectrum analysis is the most commonly
used method, which can be used to estimate the blade
frequency, propeller shaft frequency, and their corresponding
harmonics. Processing the envelope of the modulated signal
can achieve effective target recognition. Mathematical model
of the single frequency modulation signal can be described as:
2016 Sixth International Conference on Instrumentation & Measurement, Computer, Communication and Control
978-1-5090-1195-7/16 $31.00 © 2016 IEEE
DOI 10.1109/IMCCC.2016.195
19