TELKOM
NIKA
, Vol. 11, No. 9, September 20
13, pp.
5037
~50
4
3
ISSN: 2302-4
046
5037
Re
cei
v
ed Fe
brua
ry 22, 20
13; Re
vised
May 31, 20
13
; Accepte
d
Ju
ne 14, 201
3
Resear
ch on D-S Evidence Reasoning Improved
Algorithm based on Data Association
Xing Liu
*
, Sh
oushan Jia
n
g
Schoo
l of Mechan
ical En
gi
ne
erin
g, North
w
e
s
tern Pol
y
tech
nical U
n
iv
ersit
y
, Xi’a
n, 710
07
2
,
Shaan
xi, Ch
in
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: hg
yao
1
9
96@
163.com
A
b
st
r
a
ct
In order
to
make fu
ll
use
of detecti
on
info
rmati
on, i
m
pro
v
e the
anti-
int
e
rferenc
e a
b
il
ity of t
h
e
system
, s
o
lve the
problem
of target
r
e
cognition
in comple
x
env
ironment of
mult
i-s
ensor
correlation
detectio
n
, this
pap
er p
u
ts forw
ard a
infor
m
ati
on
fus
i
o
n
meth
od betw
e
en
data ass
o
ciati
on and
D-S
evid
entia
l reas
oni
ng. It discus
s
ed the co
nditi
on an
d the
me
t
hod of d
a
ta as
sociati
on, on th
e basis
of relat
e
d
infor
m
ati
on
extraction, th
e as
sign
ment
pr
ob
abil
i
ty of
mu
lti-
source s
ens
or
s was revised, it also
establis
hed
the D-S evid
e
n
tial re
ason
in
g
algor
ith
m
for target
recogn
ition b
a
sed
on
data
associ
ation.T
h
ro
ug
h th
e
establ
ish
m
e
n
t of mo
del s
i
mul
a
tion,
pr
oved t
hat the inf
o
rmation fusi
on method betw
een
data
assoc
i
ati
o
n
and D-S evi
d
e
n
tial reas
on
ing
evide
n
ce rea
s
oni
ng has
h
i
gher rel
i
a
b
il
ity and recog
n
iti
on ab
ility than
th
e
traditio
nal ev
id
entia
l reaso
n
i
n
g met
hod, the r
e
sults ve
rify th
e correctness
and effectiv
ene
ss of this meth
od.
Ke
y
w
ords
:
inf
o
rmatio
n
fusio
n
, data assoc
i
a
t
ion, D-S ev
id
e
n
ce reas
oni
ng
meth
od, targ
et recog
n
itio
n
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Information f
u
sio
n
is al
so
kno
w
n
a
s
data fu
sion,
it refers to t
he u
s
e
of compute
r
techn
o
logy fo
r the info
rma
t
ion processi
ng p
r
o
c
e
ss,
whi
c
h a
c
cord
ing to the ti
me sequ
ence to
analysi
s
and
synthesi
z
e the observati
on informat
io
n obtained u
nder
ce
rtain rule
s in ord
e
r
to
compl
e
te the
task of
de
ci
sion
-ma
k
in
g
and
estimatio
n
[1-2]. The
singl
e p
o
int i
n
formatio
n a
bout
the object and the environ
ment obtained by a plurali
t
y of s
ensors will get more compl
e
te and
reliabl
e outp
u
t throug
h the comp
reh
ensive a
nd
analysi
s
of
approp
riate fusio
n
algo
rit
h
m.
Therefore, th
e co
re
pro
b
l
e
m of multi
-
se
nsor
dete
c
tion
system
is info
rmati
on fusi
on. T
he
comm
only u
s
ed al
gorith
m
for multi
-
sen
s
or
data fu
sio
n
a
r
e
wei
ght
ed ave
r
ag
e
method, Kal
m
an
filtering meth
od, neural net
work, eviden
ce rea
s
oni
ng, fuzzy logic the
o
ry and so on
.
For
multi-se
nso
r
system,
the info
rma
t
ion
ha
s th
e
ch
aracte
risti
c
s of
diversi
t
y and
compl
e
xity. Fusio
n
of m
u
lti-se
nsor targ
e
t
re
cogniti
o
n
i
s
attem
p
ting t
o
fuse the
inf
o
rmatio
n that
is
impre
c
i
s
e a
n
d
incomplete
about the ta
rget attrib
ute
of each sen
s
or,
pro
d
u
c
in
g more a
c
curate
and complet
e
attribute estimation and
judgme
n
t
than the single
sen
s
o
r
. In many data fusi
on
method, evid
ential re
aso
n
i
ng is
suitabl
e for t
he fu
sion
without
prio
r inform
a
t
ion [2-6]. The
advantag
e of uncertainty repre
s
e
n
tation
, meas
u
r
em
e
n
t and combi
nation ha
s b
een d
r
a
w
n wi
de
attention. The traditional
evidence reasonin
g
me
thod is just
merge the
basic p
r
ob
ability
assignm
ent f
unctio
n
a
c
cording to
D-S
eviden
ce
rea
s
oni
ng m
e
tho
d
on
the fo
u
ndation
of b
a
sic
probability assi
gnment function, whi
c
h dose
not consider the inform
ation correlation
c
h
ar
ac
te
r
i
s
t
ics
b
e
t
w
e
en
mu
ltip
le
se
ns
o
r
s
[5
]. H
o
w
e
ve
r
,
in
man
y
pr
a
c
tic
a
l ap
p
l
ic
a
t
io
ns
, th
er
e is
some
correla
t
ive informati
on b
e
twe
en
variou
s
sen
s
ors, i
n
the
p
r
esen
ce
of strong
co
mple
x
environ
ment interferen
ce, the
in
ciden
ce
rel
a
tion
of target i
denti
f
ication
co
ntribution i
s
oft
en
greate
r
tha
n
the outp
u
t of singl
e sen
s
or itself .Yet,
this pa
rt of information is not
reflecte
d in th
e
traditional ev
iden
ce re
aso
n
ing, whi
c
h
doe
s not
ma
ke full use
of multi-so
urce informatio
n
.
Therefore, in
some
occa
si
ons, it is ne
e
d
to
con
s
tru
c
t a D-S evide
n
ce
rea
s
oni
n
g
method b
a
s
ed
on d
a
ta a
s
so
ciation, to
en
han
ce th
e
system
ca
pab
ili
ty of multi-source informat
ion processi
ng
and imp
r
ove the ability of target recognit
i
on.
Data a
s
soci
a
t
ed with the
D - S eviden
ce rea
s
onin
g
information f
u
sio
n
metho
d
s
mainly
studie
s
three
asp
e
ct
s. First is the
con
d
ition
of info
rmation
asso
ciated
with
D - S evide
n
ce
rea
s
oni
ng "; Second i
s
th
e method of
correl
ation
inf
o
rmatio
n acq
u
isition; Thi
r
d
is the method
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 503
7 – 5043
5038
that how
co
rrelation i
n
formation an
d
D-S evid
en
ce re
aso
n
ing
combi
ned. In
the appli
c
at
ion
backg
rou
nd o
f
target recog
n
ition, this
pa
per stu
d
ied th
e three a
s
pe
cts above.
Figure 1. Information Fu
sio
n
based on
Data Associati
o
n
2.
Analy
s
is of Informa
t
ion
Fusion
Me
thod b
e
t
w
e
e
n
Data
Asso
ciation
and
D-S Ev
idence
Rea
s
oning
2.1. D-S ev
id
ence re
asoni
ng method
Let
Z
be a set
of assu
mpti
ons, in this
collectio
n, assuming all ele
m
ents a
r
e m
u
tually
indep
ende
nt
and
com
p
lete
. The
set of a
ll sub
s
et
s of
Z sig
n
a
s
,
is
als
o
k
n
ow
n as th
e
fr
a
m
e
of disce
r
nm
e
n
t. If the
Z
ha
s n hypoth
e
si
s, then
has
n
2
sub
s
et
s,
wh
ere
is the
empty
se
t.
The eme
r
g
e
n
c
e of an evid
ence ca
n su
pport
some
s
ubsets in a
certain ex
tent .Therefo
r
e,
for
each of the
e
v
idence, there is
a
ba
si
c p
r
oba
bility assi
gnment fu
ncti
on
W
. W is
a m
appin
g
from
to
]
1
,
0
[
, als
o
satis
f
y:
0
)
(
W
;
()
1
A
WA
Whe
r
e
)
(
A
W
indicate the b
a
si
c pro
babili
ty
numbe
r of
A
,
1
)
(
0
A
W
[
6
].
)
(
A
W
is the
ba
si
c
prob
ability nu
mber only p
r
o
v
ided to
A
, refl
ects the
co
nfiden
ce
of
A
,
but it is not
A’s total co
nfiden
ce. In
order to o
b
tain
the total co
n
f
idence of
A
, it must a
dd all
sub
s
et
s B
of
A
’s basic probability numbers, us
ing the confidence function (
Bel
) expre
s
sed:
()
()
,
BA
Be
l
B
W
B
A
,
(1)
The definition
of the likeliho
od functio
n
Pl
as
follows
:
()
1
(
)
(
)
,
AB
Pl
A
B
e
l
A
W
B
A
,
(2)
)
(
A
Bel
is suppo
rt for th
e
A
, so the val
ue of the li
ke
lihood fu
nctio
n
is exp
r
e
s
sed no
doubt deg
ree
of
A
, namely,
the
estimati
on of
A
trust
value.
)]
(
),
(
[
A
Pl
A
Bel
is
call
ed the
confid
en
ce in
terval of
A
.
In pra
c
tical p
r
oble
m
s, it often appe
ars a sit
uation th
at there are several evide
n
ce
s to
sup
port a hy
pothe
sis o
r
its neg
ation. At this
moment
, it needs to cal
c
ulate the
value of
W
and
the
Bel
unde
r the
combin
ed ev
iden
ce, the st
ructu
r
al rule
s are a
s
follows:
12
12
()
0
,
()
(
)
()
()
(
)
XY
A
W
A
WX
W
Y
WA
W
A
W
A
K
Φ
(3)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Re
sea
r
ch on
D-S Evid
en
ce
Rea
s
oni
ng Im
prove
d
Algorithm
base
d
on Data
… (Xi
ng Liu)
5039
Here,
12
1(
)
(
)
XY
K
WX
W
Y
2.2. Analy
s
is
on Data As
s
o
ciation Me
thod
D-S
eviden
ce
rea
s
o
n
ing
m
e
thod i
s
ba
sed o
n
the
det
ection
of in
d
epen
dent
se
nso
r
s. I
n
many practi
cal appli
c
atio
n
s
, althou
gh t
he p
r
oba
bility assi
gnm
ent
function
of
each sen
s
or is
indep
ende
nt,
and th
ey may
dete
c
t the ta
rget at t
he
sa
me time, b
u
t
their info
rmati
on related
to
a
certai
n. This
informatio
n is not reflecte
d
in
the traditional evide
n
ce rea
s
oni
ng, whi
c
h do
es n
o
t
make full
use of multi-so
urce informa
t
ion.
The e
s
tablishm
ent
of data asso
ciation m
e
th
od
betwe
en the
sen
s
o
r
s will
make full u
s
e of th
e rel
a
ted info
rmat
ion, in o
r
de
r to improve
the
system's target recogni
tion probability [7-12].
The physi
cal
or logi
cal sy
stem desig
n make
s
the outp
u
t of multi-so
urce se
nsors exist a
certai
n rel
a
tionship, whi
c
h is mainly
manifeste
d
in
th
e
time
do
ma
in
s
p
ec
ific
s
y
nc
hr
o
nou
s
or
asyn
chrono
u
s
lo
gic,m
a
y
also
di
splay
to sp
atia
l dat
a
pa
rticular
relevan
c
e or waveform sp
ace
data, espe
ci
ally in the
detectio
n
system
of multi-source
heter
ogen
e
ous se
nsors,
the
cha
r
a
c
t
e
ri
st
ic
s of
t
he
sa
m
e
t
a
rg
et asso
ciation i
s
d
e
termin
ed by
t
he phy
sical p
r
ope
rtie
s of t
he
hetero
gen
eo
us
sen
s
o
r
. Th
is relation
ship
can
be o
b
tai
ned throug
h theoretical
de
rivation, but al
so
throug
h th
e
pre
-
sim
u
latio
n
o
r
te
st. Th
ere
are m
a
n
y
relate
d fa
ctors in
the
co
rrel
a
tion
of m
u
lti-
sou
r
ce
sen
s
o
r
s’
output,
so
me facto
r
s a
ppea
r to
su
p
port the
targ
et re
cog
n
ition
,
but othe
rs
are
oppo
site.
Ho
w to
sel
e
ct t
he
correl
atio
n facto
r
i
s
a
key
pro
b
lem
to be
solve
d
in
engin
e
e
r
ing
pra
c
tice
[8-9]
.
The outp
u
t of multi-sou
r
ce se
nsors m
a
inly incl
ude
s image,
wave
form, data
an
d
other i
n
form
a
t
ion, whi
c
h e
x
ist many correlation
fa
cto
r
s i
n
it. Acco
rding
to the
desi
gn of a
c
t
ual
detectio
n
sy
stem, the
selectio
n of related fa
ctors mu
st be
the mo
st be
neficial to t
he
determi
nation
of the target, and can
eli
m
inate all ki
n
d
s of interfe
r
ence furthe
st
. Effect of multi-
sen
s
o
r
a
s
so
ciation a
naly
s
is i
s
a
co
mplex pro
b
le
m, which ne
eds to a
naly
z
e the
spe
c
ific
probl
em
s in p
r
acti
ce.
In practi
cal a
pplication, for each sen
s
or
and
the asso
ciated sen
s
o
r
can sel
e
ct a spe
c
ific
relation
al va
ri
able val
ue
as a q
uantitative ind
e
x
of th
e
inform
ation
asso
ci
ation, t
he in
dex
can
be
cal
c
ulate
d
also can b
e
obt
ained throug
h pra
c
tical
test [10-1
6
]. A
sele
ction p
r
in
ciple of the in
dex
is when t
he
asso
ciated v
a
riabl
e rea
c
h
e
s to th
e val
ue, the target
identificatio
n
pro
bability is not
less tha
n
the
singl
e
sen
s
o
r
’
s
d
e
ci
sio
n
p
r
obability.
Th
rough
the
ratio
of calculation
of the
practi
ca
l
asso
ciated v
a
riabl
es val
u
e and the
sp
ecific
asso
ci
ated varia
b
le
s value to a
s
certain th
at the
contri
bution
of target d
e
termin
ation a
bout the
co
rrelation
information, the
ratio i
s
calle
d
correl
ation [1
1-12]. By
cal
c
ulatin
g the
degree
of
co
rrel
a
tion, to
revise th
e ta
rget recogniti
on
assignm
ent
prob
ability of
the sen
s
o
r
and to
refl
ect the
rel
a
ted info
rmatio
n of the
se
nso
r
identificatio
n appe
ars to su
pport o
r
refut
e
.
Figure 2. The
Corre
c
tion M
e
thod of Targ
et Recognitio
n
Proba
bility
Set a detection system
Z
with
P
sen
s
ors, the detection p
r
obab
ility assi
gnment of ea
ch
s
e
ns
or
to
targ
e
t
is
i
W
. The
sensor
i
P
and
)
1
(
p
q
q
sensors a
s
so
ciated. So, when eve
r
y
asso
ciation e
v
idence app
e
a
rs,
will give
sup
port to th
e se
nso
r
i
P
of target detection probability
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 503
7 – 5043
5040
assignm
ent.
Let the ratio
of co
rrel
a
tion
informatio
n
of real
mea
s
ured
rel
e
van
c
e and
the
set
t
in
g
relevan
c
e to
be
i
K
, at the same time ,setti
ng the imp
a
ct
factor of
rela
ted inform
atio
n to be
i
λ
.
Acco
rdi
ng to the releva
nt information,
it can give out the revisio
n
value of
i
W
.
The modifie
d
cal
c
ulatio
n method on
i
W
.
1
1
)
1
(
'
p
i
i
W
i
K
i
i
W
λ
,
(4)
Whe
n
i
P
does n
o
t correlate
with another
se
nso
r
,
0
i
K
'
i
W
is th
e
revi
sion
value of
i
W
. Sensor m
odified detection
probability assignment
stil
l
meets the ap
plicatio
n con
d
i
tions of D-S
eviden
ce re
a
s
oni
ng metho
d
, can be syn
t
hesi
z
ed by D-S
eviden
ce re
a
s
oni
ng synth
e
si
s method.
3. Algorithm
Design
Let
Z
be a set
of assu
mpti
ons, in this
colle
ctio
n, hypotheses el
em
ents a
r
e mut
ually
indep
ende
nt and
compl
e
te.All the coll
ection
of su
bset
s of
Z
be marked
. If
Z
has n
hypothe
sis, then
has
n
2
sub
s
et
s.
Set
Z
has p
su
pportive evid
ence, for ea
ch of
the evid
ence, there i
s
a basi
c
p
r
ob
ability
assignm
ent functio
n
W
.
W
is a mapping fro
m
to [0,1],
als
o
satis
f
ied:
()
0
,0
(
)
1
()
1
A
W
WA
WA
φ＝
＝
(5)
Suppo
sed, a detectio
n
sy
stem
Z
with P
sensors, the dete
ction probability
assignment
of each sen
s
or to target is
i
W
. The senso
r
i
P
and
)
1
(
p
q
q
senso
r
s a
s
soci
ated
. Let the
ratio of correl
ation informat
ion of re
al m
eas
ure
d
relevance an
d the
setting relev
ance to be
i
K
,
at the same t
i
me ,setting t
he impa
ct factor of relate
d informatio
n
to be
i
λ
. Ac
cording to the
relevant info
rmation, it can
give out the revision value
of
i
W
, the value
is
'
i
W
.
Similarly,
'
W
is al
so a map
p
ing
from
to [0,1], and satisfy the D-S evid
e
n
ce rea
s
oni
ng
con
d
ition
s
, it can be
cal
c
ulated a
c
cording to t
he type synthe
si
s of eviden
ce com
b
inatio
n of
)
(
'
A
W
.
Figure 3. The
Information
Fusio
n
Algorit
hm of Data Associ
ated Co
mbined
with
D-S Evidence Rea
s
o
n
ing
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Re
sea
r
ch on
D-S Evid
en
ce
Rea
s
oni
ng Im
prove
d
Algorithm
base
d
on Data
… (Xi
ng Liu)
5041
4.
Test
Resul
t
and An
aly
s
is
In orde
r to e
x
amine the f
u
sio
n
effect o
f
the D-S evi
den
ce rea
s
o
n
ing meth
od
based o
n
data asso
cia
t
ion in targe
t
reco
gnition,
it has carri
ed on the
e
x
perime
n
tal cal
c
ulatio
n a
nd
comp
ari
s
o
n
.
Let the targ
e
t
reco
gnition
frame
w
ork
as
}
3
,
2
,
1
{
O
O
O
,
3
,
2
,
1
O
O
O
are three
targets.
T
h
e
s
y
s
t
e
m
us
es
3
k
i
nd
s
of
s
e
ns
or
s
3
,
2
,
1
μ
μ
μ
, the basi
c
probability assignment of a
sampli
ng
perio
d co
rrespondi
ng a
s
shown in the matrix
D
.
D
is
expresse
d
as
follows
.
The ro
ws of the matrix co
rre
sp
ond to
sen
s
o
r
3
,
2
,
1
μ
μ
μ
, the
colum
n
s
co
rresp
ond to
,
3
,
2
,
1
O
O
O
, each
elem
ent of the
matrix rep
r
e
s
ent
the
co
rrespon
ding t
a
rget d
e
tecti
on
prob
ability of
the
corre
s
p
ondin
g
sen
s
or,
indicate the basi
c
pr
obability assi
gnment of
uncertain p
r
o
positio
n.
If we do
not
consi
der the
correlation i
n
fo
rm
ation
between the
sen
s
ors,
acco
rdin
g to D-S
eviden
ce re
a
s
oni
ng synth
e
si
s met
hod
dire
ctly, the
eviden
ce re
sult
s we can get is [0.36
0.28
0.22 0.14]. By comparing the
synt
heti
c
result and t
he det
ection
probability of
singl
e sensor, i
t
can be seen that, the
combinat
ion evi
dence
results of the
re
cognition probability obtained in
accordan
ce
with the D-S evidence rea
s
oni
ng
synthesi
s
meth
od wa
s imp
r
oved tha
n
the
detection probability of single
sensor to some extent.
In this p
ape
r, usin
g the
propo
sed
algo
ri
thm of D-S e
v
idence rea
s
oning
ba
sed
on data
asso
ciation to
do data fusio
n
cal
c
ulatio
n.
Set up th
at the two g
r
ou
ps
of sen
s
ors
3
,
2
,
1
μ
μ
μ
exist p
a
irwise a
s
so
ciat
ion, by th
e
analysi
s
of actual acq
u
isitio
n sign
al, the correl
ation ob
tained a
s
follows.
Table 1. Co
rrelation bet
we
en
a
Sensor Evidence Dist
ribution
Selecting the impact factor
2
.
0
, acco
rdin
g
to the form
ul
a (3
), fusi
ng
asso
ciated
information, calcul
ating the re
vised probability for each sensor
assi
gnment. Results are the
E,
express
ed
as
follows
.
4
.
0
2
.
0
15
.
0
25
.
0
4
.
0
15
.
0
25
.
0
2
.
0
25
.
0
2
.
0
25
.
0
3
.
0
D
258
.
0
246
.
0
185
.
0
311
.
0
265
.
0
178
.
0
306
.
0
251
.
0
085
.
0
233
.
0
293
.
0
389
.
0
E
Correl
a
ti
on
1
O
2
O
3
O
12
K
0.82
0.51
0.46
13
K
0.67
0.35
0.37
21
K
0.45
0.37
0.62
23
K
0.83
0.75
0.61
31
K
0.65
0.49
0.76
32
K
0.57
0.68
0.39
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 503
7 – 5043
5042
Comp
ared D and E,we can se
e that, after intr
od
uced the correl
ation inform
a
t
ion, the
detection probability of each
sens
ors i
s
improved than those bef
or
e introduci
ng, at the
same
time the uncertainty of the sen
s
o
r
de
crea
se
s.
This
sho
w
s that, whe
n
intro
d
u
c
ed the
relev
ant
informatio
n, the system
ca
n analyze the
target det
ect
i
on informatio
n of the multi-so
urce
sen
s
or
netwo
rk from
multiple pe
rspective
s, the target
re
co
gni
tion ability is enha
nced, su
ch a
s
FIgur 4.
Figure 4. The
Compa
r
i
s
on
betwe
en Ori
g
inal A
ssi
gnm
ent Proba
bility and Asso
ci
ation Revi
sed
Targ
et Re
cog
n
ition Prob
abi
lity
The probabilit
y assignm
ent
after
correcti
o
n
E
, accordi
n
g
to the
D-S
e
v
idence rea
s
oning
synthe
sis
me
thod , the ev
iden
ce results we
can g
e
t [0.4553 0.3
053 0.2
214
0.0180].
E
is
express
e
d
as fellows.
258
.
0
246
.
0
185
.
0
311
.
0
265
.
0
178
.
0
306
.
0
251
.
0
085
.
0
233
.
0
293
.
0
389
.
0
E
Table 2. Co
m
pari
s
on b
e
tween the O
r
igi
nal
D-S Synthesi
s
Results and the Co
rrelative
Introdu
ced Inf
o
rmatio
n Re
sults
Target identification
1
O
2
O
3
O
Without introducing Association
0.36
0.28
0.22
0.14
Introducing Association
0.4553
0.3053
0.2214
0.0180
The detection system of
correl
ation int
r
odu
ced information detect
i
on probability of the
target A increased from
0
.
36 to 0.4553
, the detec
tio
n
pro
bability of B incre
a
se
d from 0.28 t
o
0.3053, the
detectio
n
p
r
o
bability of C increa
sed
from 0.2
2
to
0.2214. By
com
pari
ng
the
synthe
sis
re
sults of the D-S eviden
ce
rea
s
oni
ng
wh
ich not int
r
od
uce
d
rel
e
van
c
e info
rmatio
n:
obviou
s
ly, the synthe
sis
result of the
correl
atio
n
introdu
ce
d in
formation of
D-S evide
n
c
e
rea
s
oni
ng h
a
s
hig
h
e
r
re
co
gnition p
r
ob
a
b
ility and a
c
cura
cy. At the same
time, th
e un
ce
rtainty of
the syste
m
is redu
ce
d to 0
.
0180 fro
m
0.
14. It mean
s
that, making
full use of vari
ous i
n
form
ation
of
mult
i-
sou
r
ce se
ns
or
s,
while
maintai
n
ing th
e h
a
rdwa
re
of th
e ori
g
inal
de
tection
syste
m
unchan
ged, the data fusi
on method
can gre
a
tly im
prove the
detectio
n
performan
ce of
the
system, such as Figu
re 5.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Re
sea
r
ch on
D-S Evid
en
ce
Rea
s
oni
ng Im
prove
d
Algorithm
base
d
on Data
… (Xi
ng Liu)
5043
Figure 5. Co
mpari
s
o
n
the Origin
al D-
S Synthesis a
n
d
Correl
ative Introdu
ced
Information
5. Conclu
sion
As can b
e
se
en from the simulation re
sult after introduci
ng the co
rrel
a
tion information,
by revi
sing the original
probab
ility assi
gnment, the accuracy
of
D-S
evidence reasoni
ng
method
has be
en i
m
p
r
oved
obviou
s
ly. And th
e
uncertainty of
the
system
i
s
g
r
e
a
tly red
u
ce
d. By flexibly
usin
g this m
e
thod, it can
also solve t
he anti-
inte
rf
eren
ce p
r
obl
em of particular environ
ment
targeted. T
h
is pape
r p
r
ovid
es a
ap
proxi
m
ate ap
pr
o
a
ch for e
ngin
e
e
r
ing
pro
c
e
s
si
ng an
d a t
r
ai
n of
thought fo
r correlating i
n
formatio
n bet
wee
n
sen
s
or
s which is a
compl
e
x issu
e asso
ciate
d
with
desi
gning
the
system
structure in
practi
ce. Ho
w to
effectively
use multiple sen
s
or
info
rmatio
n
to
detect sy
ste
m
need
s furth
e
r re
se
arch a
nd exploratio
n.
Referen
ces
[1]
Xi
an
g -
y
i Ch
en
, F
e
i Li, Lia
ng-
min W
ang.
F
a
ult-T
o
ler
ant Method for D
e
tecting C
o
ver
a
ge Ho
les
o
f
W
i
reless Sens
or Netw
orks
.
TELKOMNIKA Indo
nesi
an Jo
u
r
nal of
Electric
al Engi
ne
erin
g
.
2012; 10 (4):
876-
882.
[2]
Z
hou Bing-
yu,
Lu Ye, Li Z
h
u
a
ng, W
u
Hai-n
i
n
g
.
D-S Evidenc
e theory
meth
o
d
in the a
ppl
ica
t
ionof targ
et
recog
n
itio
n
.
Ph
otoel
ectric tech
nol
ogy a
ppl
icat
ion
. 20
07; 22(
5
)
: 67-70.
[3]
Hans
han
Li, Z
h
i
y
o
ng L
e
i, Z
e
min W
ang, Ju
ncha
i Gao.
Re
search Obj
e
ct Photoe
lectric
Char
acteristi
c
and F
i
r
e
Co
ord
i
nate
Distrib
uti
ng Pro
b
a
b
il
it
y
i
n
Across Scre
e
n
S
y
st
em.
Jour
nal
of Nan
o
e
l
e
c
tronics a
n
d
Optoelectro
n
ic
s.
2012; 7(2): 1
99-2
03.
[4]
W
e
iming
Ya
ng
, Jian
Z
han
g, J
i
n
x
ia
ng
Pe
ng.
Mean-
s
quar
e
Error Distorti
o
n
Comp
utatio
n
Mode
l B
a
se
d
on W
i
re
less S
e
nsor N
e
t
w
ork.
T
E
LKOMNIKA Indon
esi
an J
o
u
r
nal
of El
ectrica En
gin
eeri
n
g
. 201
2;
10(
7)
:
172
3-17
29.
[5]
Lia
o
C
h
o
ngd
a, Gao Gu
i-min
g
.
D-S Evi
denc
e the
o
ry
in
th
e a
ppl
icati
on
of rad
a
r targ
et
reco
gniti
on
.
Rad
a
r and
aga
inst. 2008; 3(
3)
: 32-34.
[6]
Xi
aofa
n
W
u
, Chun C
hen, Ji
aj
un Bu. An Inte
rf
erence-
a
w
ar
e
T
r
ansmission Po
w
e
r
C
ontrol Scheme
for
Vehic
u
lar
On-
boar
d Mo
nitor
i
ng W
i
r
e
less
Sensor
Net
w
o
r
k.
T
E
LKOMNIKA Indo
nesi
a
n Jo
urn
a
l
of
Elecrica
l Eng
i
n
eeri
n
g
. 20
12; 1
0
(5): 107
7-1
0
8
0
.
[7]
Hans
han
Li, Z
h
i
y
o
ng
Lei, Z
e
min W
a
n
g
, Jun
c
hai Ga
o.
Res
earch
on Ob
je
ction Infor
m
ati
on Extractio
n
Arithmetic i
n
P
hoto -
e
lectric
D
e
tection
T
a
rg
et Base
on
W
a
v
e
let A
nalys
is
Method
. Prz
egl
ad
Elektrote
-
chnicz
n
y
. 2
0
1
2
;
88(9): 157-1
6
1
.
[8]
F
u
Yan-ju
n, Su
n Kai-fen
g
, Shi
Chao.
D-S Ev
ide
n
ce the
o
ry and its a
ppl
icat
ion i
n
target re
cogn
ition
. Air
F
o
rce eng
in
eer
ing u
n
ivers
i
t
y
j
ourn
a
l. 20
08; 9
(
4): 50-52.
[9]
Xu J
un-
ya
n. D-S evid
ence
theor
y
of informatio
n
fusio
n
method
app
licatio
n in targ
et recogn
itio
n
.
Journ
a
l of bo
mb arr
o
w
and
gui
danc
e
. 200
5; 25(3): 84-8
7
.
[10]
Hans
han
L
i
, Z
h
i
y
o
ng
Le
i.
Pr
ojectil
e
T
w
o-di
me
nsi
ona
l C
o
ordi
nates M
e
a
s
ure
m
e
n
t Met
hod
Bas
ed
o
n
Optical F
i
ber
C
odi
ng F
i
re a
n
d
its Coord
i
nat
e
s
Distributi
ng
Proba
bil
i
t
y
.
Me
asurem
ent Sci
ence R
e
vi
e
w
.
201
3; 10(1): 34
-38.
[11]
Z
hou Bin
g
-
y
u,
Lu Ye, Li Z
h
u
a
ng, W
u
Hai-n
i
n
g
.
D-S Evidenc
e theory
meth
o
d
in
the a
ppl
ica
t
ion of target
recog
n
itio
n.
Ph
otoel
ectric tech
nol
og
y a
ppl
icat
ion. 20
07; 22(
5
)
; 67-70.
[12]
Henri
on M. Propa
gati
ng
Uncertainty in Bayesi
a
n
N
e
tw
orks by Proba
bil
i
stic Lo
gic Sa
mp
lin
g
.
Uncerta
i
nt
y
in
Artificial Intel
lig
ence. 19
98; 15
(2): 149-1
64.
[13]
Yu Y, Pras
ann
a
VK, Krishnamachari B.
En
ergy Mi
ni
mi
z
a
t
i
on for
Re
al-ti
m
e Dat
a
Gather
i
ng i
n
W
i
re
les
s
Sensor N
e
tw
orks
. IEEE
Transactions
on Wireless Communi
cations. 2006;
5(11): 3087-
3096.
[14] Mahl
er
RPS.
Co
mbi
n
i
ng A
m
b
i
g
uous
Evi
denc
e w
i
th R
e
spect to
A
m
big
uous
a Pr
i
o
ri Kn
ow
led
g
e
Bool
ean
Lo
gic
. IEEE
T
a
rans o
n
S
y
stems, Ma
n and
C
y
ber
n
Part A: Sy
stem
sand
Huma
ns. 199
6; 26(1)
:
28-4
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.