第
44
卷第
7
期
2010
年
7
月
上海交通大学学报
JOURNAL
OF
SHANGHAI
JIAOTONG
UNIVERSITY
文章编号
:1006-2467(2010)07-0873-05
Vol. 44
No.7
Ju
l. 2010
一种新的多机动目标跟踪的
GMPHD
滤波算法
郝燕玲
1
,
孟凡彬
1.2
,
王素鑫
2
,
孙枫
1
(l.哈尔滨工程大学自动化学院,哈尔滨
150001;
2.
天津航海仪器研究所,天津
30013
1)
摘
要:针对多机动目标跟踪的传统数据关联算法约束条件苛刻、估计精度低、计算量大等问题,
提出了一种基于随机集理论的非数据关联的多机动目标跟踪算法.该算法将高斯混合概率假设密
度
(GMPH
D)滤波与"当前"统计模型的优点相结合,绕过了棘手的数据关联问题,能高效处理目标
数较大的机动跟踪问题.在漏检、虚警、多机动目标交叉杂泼复杂环境下进行了仿真实验,结果表
明,该算法具有较高的跟踪精度和稳健的跟踪性能.
关键词:多机动目标跟踪;随机有限集
p
高斯混合概率假设密度滤.波
z
扩展卡尔曼滤泼
申圄分类号:
TP18;
TP274
文献标志码
:A
A New GMPHD Filter Algorithm for
Multiple Maneuvering Targets Tracking
HAO
Yan-ling
1
,
MENG
Fan-bin
1
•
2 ,
WANG
Su-xin
2
,
SUN
Feng
1
O.
College of
Automation
,
Harbin
Engineering
University
,
Harbin
150001 ,
China;
2.
Tianjin
Navigation
Instrument
Research
Institute
,
Tianjin
300131 ,
China)
Abstract:
Considering
the
traditional
data
association
algorithm
of
multiple
maneuvering
targets
tracking
being of
hard
constraint
condition,
lower
estimated
accuracy,
and
higher
computational
complexity, a
non
data
association
tracking
algorithm
based
on
the
random
set
theory
was proposed. Since
the
proposed
algo-
rithm
integrates
the
both
advantages
of
Gaussian
mixture
probability
hypothesis
density
(GMPHD)
filter
and
current
statistical
mode1 , avoids
the
difficult
problem
of
data
association,
it
is able
to
deal
with
multi-
ple
maneuvering
targets
tracking
effectively. A
simulation
experiment
was
performed
in
the
complex envi-
ronment
with
clutter
,
miss
detection
, false
alarm
,
dense
,
and
cross
targets.
The
simulation
results
show
that
the
proposed
algorithm
has
higher
tracking
accuracy
and
more
steady
tracking
performance.
Key
words:
multiple
maneuvering
tracking;
random
finite
sets;
Gaussian
mixture
probability
hypothesis
density(GMPHD)
filter;
extended
Kalman
filter
在多目标跟踪中,一些传统数据关联算法,如联
合概率数据关联
Ooint
Probability
Data
Associa-
tion
,
JPDA)
和多假设跟踪
(Multiple
Hypothesis
Tracking
,
MHT)
等虽然在某些方面有一定优势,
但随着杂波数或目标数的增加,会导致"组合爆炸"
收稿日期:
2009-06-22
蕃金项目:国家自然科学基金资助项目
(60704018)
问题,进而使算法失效[叫.
20
世纪初,
Mahler[3
J
提
出基于随机集理论的高斯混合概率假设密度
(Gaussian
Mixture
Probability
Hypothesis
Densi-
ty
,
GMPHD)
滤披的非数据关联眼踪算法,它绕过
了复杂的数据关联问题,具有计算量小、实现简单,
作者简介
z
郝燕玲(1
944-)
,女,山东烟台人,教授,博士生导师,主要研究方向
z
导航、制导与控制.
电话
(Te
l.)
:0451-82519907IE-mail.haoyanling@hrbeu.edu.cn.