Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
4060
~4
068
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp4060
-
406
8
4060
Journ
al
h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
A
h
yprid
t
echn
iqu
e for
hu
man f
ootprint
r
ec
og
niti
on
Yahya I
smail
Ibra
him
,
Isr
aa
Moham
med
A
lh
amdani
Coll
ege of
Educat
ion
for
Pure
Sc
ie
nc
es,
Dep
art
m
ent
of
Com
pute
r S
ci
ence, Unive
rs
ity
of
Mos
ul
,
Ir
a
q
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Dec
9
, 2
0
18
Re
vised
A
pr
16
, 2
01
9
Accepte
d
Apr
27
, 201
9
Biom
et
ric
s
has
c
once
rne
d
a
gre
at
ca
re
recent
l
y
du
e
to
i
ts
importan
t
in
th
e
li
f
e
tha
t
st
art
s
fro
m
ci
vil
appl
i
c
at
ions
to
s
ec
ur
ity
and
re
ce
nt
l
y
te
rror
ism
.
A
Footprint
recognit
ion
is
one
of
the
per
son
al
ide
n
ti
fi
cation
s
base
d
on
biometri
c
m
ea
su
rement
s.
Th
e
aim
of
thi
s
rese
ar
ch
is
to
design
a
prope
r
and
rel
i
abl
e
biometric
s
y
st
em for
hu
m
an
footpri
nt
r
e
cogni
ti
on
named
(FRB
S) t
hat
stands
for
F
ootpri
nt
Re
cogni
t
ion
Biom
et
ric
S
y
ste
m
.
In
addi
ti
on
,
t
o
construct
a
hum
an
footpr
int
databa
se
wh
ic
h
it
is
ver
y
hel
pful
for
v
arious
use
in
scie
ntific
appl
i
c
at
ion
e
.
g.
for
a
uthe
ntica
ti
on.
T
her
e
ex
ist
m
an
y
biometri
cs
dat
ab
ase
s
for
othe
r
ide
nt
ity
but
v
er
y
r
are
f
or
foot
print
.
As
well
as
the
exi
sting
one
ar
e
ver
y
lim
it
ed.
Thi
s
pap
er
pre
sen
ts
a
ro
bust
h
y
pr
id
t
ec
h
nique
s
whic
h
m
erg
es
bet
wee
n
Im
age
Proce
ss
ing
with
Artifi
cial
Intelli
g
ent
technique
vi
a
Ant
Colon
y
Opti
m
iz
at
ion
(ACO
)
to
rec
ognize
hum
an
footpri
nt.
(ACO
)
pl
a
y
s
the
essent
ia
l
ro
le t
hat
r
ise
the
p
erf
orm
anc
e
and
th
e
qual
ity
o
f
the
r
e
sults i
n
the
biometri
c
s
y
ste
m
via
feature
se
le
c
ti
on.
The
set
of
the
se
le
c
te
d
f
ea
tur
es
wa
s
tre
a
te
d
as
expl
o
rat
or
y
information,
and
select
s
the
opti
m
um
fea
ture
set
i
n
standi
ngs
of
fe
a
t
ure
set
siz
e.
Li
f
e
RGB
footpri
n
t
images
from
ni
ne
per
sons
with
te
n
images
per
per
son
const
ruc
te
d
from
li
f
e
visual
dataset
.
At
first,
th
e
visual
dataset
w
as
pre
-
proc
esse
d
oper
at
ions
.
E
ac
h
resul
ta
nt
image
de
te
c
ts
footpri
nt
that
is
cro
pped
to
por
ti
ons
rep
r
ese
nt
e
d
b
y
thr
ee
blocks
.
The
first
bloc
k
is
for
finge
rs,
the
sec
ond
bloc
k
ref
ers
to
t
he
ce
nt
er
of
the
foot
and
the
la
st
one
determ
ine
s
the
hee
l
.
Th
en
fea
tur
es
were
ext
racte
d
from
ea
ch
imag
e
and
stored
in
Exc
el
file
to
be
entere
d
to
Ant
Colon
y
Optimization
Alg
orit
hm
.
The
exp
eri
m
enta
l
outc
om
es
of
th
e
s
y
stem
show
t
hat
th
e
proposed
al
gori
thm
eva
lu
at
es
opt
imal
result
s
with
sm
al
le
r
feature
set
compari
ng
with
othe
r
al
gorit
hm
s.
Exp
eri
m
ent
a
l
out
co
m
es
show
tha
t
our
al
gor
it
hm
obta
ins
an
eff
icient
and
acc
ura
te
result
abo
u
t
100%
a
cc
ur
a
c
y
in
compari
so
n
with
oth
er
rese
arc
h
es
on
th
e
sam
e
f
ie
ld
.
Ke
yw
or
d
s
:
An
t
c
olony
o
pt
i
m
iz
ation
(
AC
O
)
B
iom
e
tric
F
eat
ur
es
selec
ti
on
Footp
rint
r
ec
ogniti
on
Mor
phologica
l fil
te
r
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Isr
aa
Mo
ham
med Alham
dan
i,
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nces,
Un
i
ver
sit
y o
f
Mosu
l,
Mosu
l,
Iraq
.
Em
a
il
: esra
m
k6
5
@
gm
ail.co
m
1.
INTROD
U
CTION
Re
li
able
per
s
onal
rec
ogniti
on
ha
s
bee
n
a
n
i
m
po
rtant
c
om
pu
te
r
app
li
ca
ti
on
.
S
om
e
exa
m
ples
wer
e
est
ablishin
g
a
pe
rson
ide
ntit
y
[1
]
.
T
he
di
gital
Im
age
processin
g
em
pl
oyed
m
at
hem
a
ti
cal
,
Stat
ist
ic
a
l
an
d
Op
ti
m
iz
ation
t
o
bio
lo
gical
sci
ence k
no
wn
as
Bi
om
et
ric
[2
]
.
Bi
om
et
rics
sy
stem
is
a
patte
r
n
re
co
gnit
ion
s
yst
e
m
that
e
m
plo
ye
d
to
recog
nize
and
i
den
ti
fy
a
per
s
on
base
d
on
ph
ysi
ol
og
ic
al
char
act
erist
ic
s
[3
]
.
Mor
eov
e
r,
the
be
hav
i
or
al
char
act
e
risti
cs
Table
1
cat
e
gorize
the
va
rio
us
ty
pe
of
bio
m
et
rics.
The
footp
rints
rec
ogni
ti
on
wer
e
one
of
t
he
im
po
rtant
form
s
of
bi
om
et
ric
identific
at
ion
us
ed
f
or
La
w
an
d
ci
vili
an
ap
plica
ti
on
[4
]
Bi
om
e
tric
s
in
gen
e
ral
hel
ps
to
recog
ni
ze
a
per
s
on
accor
ding
t
o
physi
og
nom
ie
s
or
trai
ts
[5
]
.
The
physi
og
nom
ie
s
su
ch
as
s
ign
at
u
re,
gait
or
pass
word,
wh
il
e
trai
t
refe
rs
to
the
str
uc
tural
patte
r
n
su
c
h
as
fin
gerpr
i
nts,
f
oot,
ear
,
face
,
DNA,
de
ntal
ti
ssu
e
[5
]
.
F
ootpri
nt
bio
m
et
rics
co
uld
be
ap
pl
ie
d
us
ef
ully
in
holy
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A h
y
pr
id
te
ch
ni
qu
e f
or
hum
an
foo
t
pr
int rec
ogniti
on (Y
ah
y
a Is
ma
il
Ib
r
ahim)
4061
places
w
here
s
ho
e
s
rem
ov
e
d
inh
e
ren
tl
y
due
to
the
cr
owed
.
Pers
on
ide
ntif
ic
at
ion
,
ne
wborn
ide
ntific
at
ion
t
o
protect
infant
f
ro
m
lost. Ch
ec
k po
i
nt at ai
r
port [2]
.
Table
1.
Bi
ome
tric
Type
with
Tr
ai
ts a
nd thei
r
U
sage
[2]
Bio
m
etric
Feature
Fo
r
m
of
Bio
m
et
ric
Deter
m
in
ed
Easin
ess
of
us
e
Fin
g
erprint
Ph
y
sical
Hig
h
Mediu
m
Fo
o
tp
rint
Ph
y
sical
Un
k
n
o
wn
Low
Pal
m
Print
Ph
y
sical
Hig
h
Mediu
m
Ir
is
Ph
y
sical
Hig
h
Low
Vo
ice
Beh
av
io
r
Mediu
m
Hig
h
Face
Ph
y
sical
Low
Hig
h
R
edu
ct
io
n
of
patte
rn
dim
en
sion
al
it
y
ref
er
s
to
featur
e
s
el
ect
ion
thr
ough
f
eat
ur
e
e
xtracti
on
th
i
s
process
play
s
a
c
onsidera
bl
e
im
po
rtance
m
issi
on
f
or
pa
tt
ern
recog
niti
on
[
6].
T
he
obj
ect
ive
of
f
eat
ur
e
sel
ect
ion
is
to
el
i
m
inate
un
ne
cessary
featu
re
s
and
sel
ect
av
ai
la
ble
featur
es
.
The
know
n
al
gorithm
An
t
C
olony
Op
ti
m
iz
ation
(ACO
)
is
us
e
d
f
or f
eat
ure sel
ec
ti
on
with
heuri
sti
c search t
hat
r
e
du
ce
the c
om
pu
ti
ng
ti
m
e [
7].
Du
e
to
the
im
po
rta
nce
of
f
ootpri
nt
rec
ogniti
on
as
disc
us
se
d
a
nd
it
is
rar
e
ly
us
ed
f
ootpri
nt
bi
om
et
ric
syst
e
m
s
in
var
iou
s
est
ablis
hm
ent.
This
pa
per
pr
opos
e
d
a
fo
ot
pr
i
nt
biom
et
ric
syst
e
m
and
c
onstruct
f
oo
t
pr
i
nt
database
im
ag
es.
T
her
e
exist
m
any
bio
m
et
r
ic
s
databa
ses
f
or
ot
her
ide
ntit
y
su
c
h
as
face
iris
but
ver
y
ra
re
f
or
footp
rint. As
w
el
l as the e
xisti
ng one a
re
v
er
y l
i
m
it
ed.
W
e
s
olv
e t
his
prob
l
e
m
b
y coll
ect
ing l
ife RGB
f
ootp
rint
i
m
ages
fr
om
nin
e
per
s
on
with
te
n
i
m
age
ea
ch
the
total
is
90
f
ootp
rint
im
age.
The
ne
w
idea
with
this
syst
e
m
was
t
he
c
om
pen
sat
io
n
betw
een
im
age
pro
cessi
ng
te
ch
ni
qu
e
s
a
nd
the
intel
li
gen
t
m
eth
od
via
A
nt
Colo
ny
Op
ti
m
iz
ation
A
lg
or
it
hm
(AC
O)
i
n
this
te
chn
i
qu
e
feat
ur
e
s
we
re
sel
ect
ed
in
a h
e
ur
ist
ic
s
trat
egy
that ob
t
ai
n
be
st
syst
e
m
per
f
orm
ance
with
be
st
op
ti
m
al
feat
ur
es
.
I
nvest
ig
at
ion
al
res
ults
on
im
age
data
set
s
disp
la
y
t
hat
the
pro
po
se
d
proc
ess
ha
s
s
uperi
or
pe
rfo
rm
ance.
Com
par
in
g
with
oth
e
r
e
xi
sti
ng
proce
dur
es,
our
al
gorithm
gain
bette
r
acc
ur
acy
with
a
sm
al
le
r
featu
re
set
.
Re
st
of
this
pa
per
is
str
uctu
red
a
s
f
ollows
r
el
at
ed
pap
e
r
t
his
s
ect
ion
discuss
t
he
li
te
ratur
e
deal
w
it
h
footp
rint
r
ecognit
ion,
A
nt
Colon
y
O
ptim
iz
at
ion
sect
i
on
desc
ribes
t
he
ide
a
about
a
nt
c
olony
a
nd
it
s
be
ha
vior,
A
nt
Col
on
y
O
ptim
iz
ation
Al
gorithm
sho
w
how
an
t
colo
ny
beh
a
vior
i
s
translat
ed
to
a
m
at
he
m
at
ic
a
l
process
.
In
a
ddit
ion
,
Mor
phol
og
ic
al
filt
er
focuses
on
ed
ge
detect
ion
f
or
f
o
ot
pr
i
nt
i
m
age,
Propse
d
work,
Visu
al
databas
e
acq
ui
sit
ion
this
part
sh
ow
how
w
e
colle
ct
li
fe
im
age,
Vis
ual
dataset
pr
e
processi
ng.
As
well
as,
footp
rint
feat
ure
extracti
on
sta
ge,
rec
ognit
ion
te
ch
nique
sta
ge,
e
xp
e
rim
ental
ou
tc
om
es, co
nc
lusio
n
a
nd Refere
nces.
2.
RELATE
D
P
APER
Ov
e
r
rece
nt
a
nd
la
st
dec
ad
e
f
oo
t
pr
i
nt,
r
e
cogniti
on
has
been
st
udie
d,
analy
zed
a
nd
researc
he
d.
The
relat
ed
w
orks
with
dif
fere
nt
and
var
i
ous
m
et
ho
ds
are
sh
ow
n
bel
ow
.
Al
-
D
ulaim
i
Reco
gniz
e
an
d
id
entify
per
s
on
dete
rm
i
nan
t
value
of
f
oo
t
pr
i
nt
after
t
he
im
age
is
se
gm
ented
to
bloc
ks
an
d
th
res
hold
is
use
d
to
extract
featur
e
s
[
4].
K
um
ar
and
D
ub
ay
us
e
m
a
inly
footp
rint
rec
og
niti
on
proce
dure
w
hich
cat
eg
or
iz
e
d
into
th
re
e
ste
ps
i
m
age
pr
e
-
pro
c
essing,
feat
ure
extracti
on
a
nd
m
at
ching
s
ta
ge.
Im
age
pre
-
proce
ssin
g
is
the
process
use
d
t
o
el
i
m
inate
the
no
ise
f
ro
m
or
i
gin
al
im
age.
I
n
f
eat
ur
e
ext
ra
ct
ion
sta
ge,
t
he
featu
res
s
uc
h
as
press
ur
e,
locat
ion
and
weig
ht
are
extracte
d
f
rom
an
i
m
age.
Indeed,
the
m
at
c
hing
proces
s
be
tween
f
ootp
rint
recog
niti
on
and
the
input
i
m
age
is
ev
al
uate
d
[
5].
Anothe
r
pap
e
r
us
e
h
y
per
pr
i
nc
iples
of
featu
r
e
extracti
on
f
uz
zy
log
ic
an
d
neural
netw
ork
[
8].
T
he
f
oo
t
pr
i
nt
is
crop
ped
a
nd
re
siz
ed
then
m
odifie
d
h
aa
r
e
ne
r
gy
featu
res
is
struct
ured
.
Ma
tc
hing
op
e
rati
on
is
pe
rfor
m
ed
v
ia
Eu
cl
idian
distanc
e
betw
een
MH
E
feat
ur
es
an
d
featur
e
s
st
or
e
d
in
t
he
databas
e
[
1].
Wh
il
e
ot
her
pa
per
us
e
geo
m
et
ry
char
act
eri
sti
cs,
sh
ape
a
nd
te
xture
[
2].
Oh
l
a
nd
Wild
determ
ines
the
m
ai
n
featur
e
s
of
the
f
oo
t.
Th
ese
f
eat
ur
es
base
d
on
ge
om
et
ry
com
pu
te
s
sh
a
pe
an
d
te
xtu
re
[
9]
they
us
e
footp
rint
sh
a
pe
that
are
base
d
on
a
sm
a
ll
nu
m
ber
of
fe
at
ur
es.
T
hey
app
li
ed
geo
m
et
ric
m
or
phom
et
r
ic
te
chn
ique
to
study
footp
rint s
ha
pe
v
a
riat
ion
of sa
m
ple o
f
83 a
du
lt
w
om
en
[
10
]
.
3.
AN
T
COL
ONY
O
PTIMIZ
A
TION
An
t
c
om
plex
beh
a
vior
ha
ve
been
st
ud
ie
d
analy
zed
an
d
e
m
plo
ye
d
in
de
sign
i
ng
fruit
f
ul
m
od
el
fo
r
eff
ect
ive
c
om
bin
at
or
ia
l
opti
m
iz
at
ion
so
luti
on
al
gorithm
s.
An
t
c
olony
optim
iz
at
ion
is
a
m
et
aheu
risti
c
wer
e
ants
c
oope
rate
in
fin
ding
go
od
s
olu
ti
ons
t
o
dif
ficult
discr
et
e
optim
iz
ation
pro
blem
s
[1
1,
12]
.
An
t
C
ol
on
y
is
der
i
vatives
of
swar
m
intel
lig
ent.
A
nt
Col
on
y
Op
ti
m
iz
ation
te
c
hn
i
qu
e
fin
d
s
olu
ti
on
to
a
m
et
a
heu
risti
c
pro
blem
s
by
a
dd
i
ng
var
ia
ble
to
const
ru
ct
a
com
plete
on
e
or
c
hange
r
epeate
dly
m
od
ify
ing
s
om
e
o
f
it
is
el
e
m
ents to
ac
hieve bet
te
r o
ne
as s
how
n
in
F
igur
e
1
[11].
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4
0
6
0
-
4
0
6
8
4062
(a)
(b)
(c)
Figure
1. (
a
) A
nts
us
e
rand
om pa
th
b
et
ween
nest a
nd food
s
ource
,
(
b
)
Dur
i
ng r
et
urn jou
rney
an
t l
eaves
behin
d
the traces
of
phero
m
on
es
,
(
c
) Phe
ro
m
on
e is
pu
t
fast
on the
sh
ort
er
p
at
h
In
it
ia
ll
y
ants
ha
ve
no
idea
of
wh
e
re
foo
d
is
i
n
the
e
nv
ir
on
m
ent,
they
sea
rch
f
or
f
ood
by
exp
l
or
e
th
e
area
s
urrou
nd
i
ng
their
nest
i
n
a
rand
om
m
ann
e
r.
As
soo
n
as
a
n
a
nt
fi
nd
s
a
f
ood
s
ource
,
it
evalu
a
te
s
the
qu
a
ntit
y and t
he
quali
ty
o
f
t
he
f
oo
d
a
nd car
ries so
m
e o
f
it
w
al
k
bac
k
t
o
the
nest.
Durin
g
t
he
r
et
urn
trip
, the
an
t
pu
ts
a
c
hem
ic
a
l
su
bs
ta
nce
cal
le
d
pher
om
on
e
on
t
he
gro
und.
An
ts
c
omm
un
ic
at
e
to
each
ot
her
by
m
ark
ing
an
d
fo
ll
owin
g
path
via
ph
e
ro
m
ones.
Ot
her
ants
are
able
t
o
sm
el
l
this
ph
e
r
omon
e
.
Its
prese
nc
e
eff
ect
s
t
he
cho
ic
e
of
t
heir
path,
that
is,
they
te
nd
t
o
f
ollo
w
s
tro
ng
pher
om
on
e
s
o
w
he
re
a
nts
go
within
and
ar
ound
th
ei
r
ant
colo
ny,
this
w
al
k
is
based
on
sti
g
m
erg
ic
syst
e
m
.
If
an
ant
decides
to
fo
ll
ow
the
trai
l
it
l
ay
m
or
e
ph
e
rom
on
es
to apply
the
tra
il
[
12, 1
3].
The
n
the
pher
om
on
es
within
a
gr
ound
hel
ps
ants
to
fin
d
the
shor
te
st
trai
ls
to
the
foo
d
source
.
So
t
hat,
ph
e
r
om
on
e
is
inv
e
rs
el
y
propor
ti
on
al
to
the
path
.
The
st
ron
g
ph
e
ro
m
on
es
t
he
s
hortest
path
t
o
t
he
f
oo
d
so
urce.
AC
O
al
gorithm
in
ge
ner
al
c
onta
in
s
a
series
of
ste
ps
s
uch
as
ph
e
r
om
on
e
ini
ti
al
iz
ation
,
c
re
at
e
a
so
luti
on a
nd
update
pher
om
on
e.
C
onti
nu
e
unti
l a te
rm
inal
conditi
on [1
3].
4.
AN
T
COL
ONY O
PTIMIZ
A
TION
ALG
O
RITH
M
Wh
e
n
a
nts
m
ov
e
f
r
om
the
nest
i.e.
point
A
to
t
he
f
ood
so
urce
i.e
.
poi
nt
B
ants
le
a
ve
a
chem
ic
al
,
p
he
r
om
on
e,
to
m
ark
these
pa
ths.
An
t
te
am
us
e
pro
bab
il
it
y
to
their
pat
h
wh
ic
h
hav
e
gr
eat
er
co
nce
ntr
at
ion
of
ph
e
r
om
on
e.
Th
e
al
go
rithm
is
base
d
on
ada
pt
ively
update
the
ph
e
ro
m
on
e
on
pat
h
at
each
node.
In
gener
al
An
t
colo
ny alg
or
it
hm
is descri
bed
br
ie
fly
in
F
ig
ur
e
2
[
6,
7, 11]
. Pseu
do
-
c
od
e
in
g
ive
n
i
n
Fi
gur
e
3
[
13
, 14].
Pro
cedu
re
ACO
A
lg
o
rith
m
I
NI
TI
L
IZ
E
an
d
List Activities
C
ONSTR
U
CT Ant So
lu
tio
n
s
DO inn
er
f
o
rm
A
ctio
n
s (op
tio
n
al)
ADJUS
T
Ph
ero
m
o
n
es
E
ND
LI
S
T
AC
T
IVITIE
S;
End
Pr
o
cedu
re
Figure
2. O
utli
ne of
Ant Col
ony Al
gorithm
(
ACO
)
Pro
cedu
re
ACO C
o
m
p
u
tatio
n
Algo
ri
th
m
Po
p
=I
n
itializa
tio
n
Esti
m
ate
Pop
u
latio
n
(
p
o
p
)
So
p
ti
m
al=
Sho
rt
p
ath
Of
Pop
u
latio
n
(
p
o
p
)
wh
ile (
cu
rr
en
t o
p
ti
m
al path
leng
th
no
t
m
et)
d
o
p
o
p
’
Rec
reate
(
po
p
)
p
o
p
’’
Mutatio
n
(
p
o
p
’)
Evalu
ate Pop
u
latio
n
(
p
o
p
’’
)
S
=Op
ti
m
a
l pat
h
Of
Pop
u
latio
n
(
p
o
p
’’
)
If
f
(
s) <
f
(
s
o
p
ti
m
al
)
th
en
Sb
est= s
en
d
-
if
p
o
p
Recr
eate(
p
o
p
’’
)
en
d
-
wh
ile
return s
h
o
rt
p
ath
Figure
3.
Pse
udo
-
c
od
e
for (
A
CO)
al
gorithm
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A h
y
pr
id
te
ch
ni
qu
e f
or
hum
an
foo
t
pr
int rec
ogniti
on (Y
ah
y
a Is
ma
il
Ib
r
ahim)
4063
5.
MO
RPH
OLO
GICAL FIL
T
ER
Mor
phologica
l
i
m
age
pr
oc
essing
is
de
rive
d
from
Ma
them
a
ti
cal
Morph
ology
(MM
)
a
nd
m
at
he
m
at
ic
a
l set
theory. T
his
te
chn
iq
ue
is
u
s
ually
u
se
d
to
de
te
ct
o
bject
bo
unda
ries (e
dge
s).
As well
, it i
s u
se
d
as
pri
m
ary
pr
ocessin
g
te
c
hniq
ue
for
thi
nn
i
ng
an
d
pr
un
i
ng
of
e
dg
es.
Ge
ne
rall
y,
m
os
t
m
or
phologica
l
op
e
rati
ons
are
base
d
on
sim
ple
sh
rin
king
an
d
ex
pandin
g
operati
ons.
T
he
two
m
os
t
co
m
m
on
m
or
phol
og
ic
al
op
e
rati
ons
a
re
ero
si
on
an
d
dilat
ion
.
In
this
w
ork
four
m
or
ph
ologica
l
filt
ers
are
use
d
these
filt
ers
are
disc
us
se
d
in
f
ollow
s
.
Cl
os
e
op
e
rati
on
is
a
dilat
ion
fo
l
lowe
d
by
a
n
e
ro
si
on.
M
orp
holo
gical
cl
os
in
g
is
use
d
t
o
jo
in
th
e
pix
el
s
in
a
n
i
m
age
tog
et
her
.
This
is
do
ne
by
fill
ing
in
the
ga
ps
betwe
en
them
and
s
m
al
l
dar
k
s
po
t
s
ar
e
rem
ov
ed
in
te
r
n
sm
oo
thin
g
th
ei
r
ou
te
r
e
dg
e
s
.
W
hile,
cl
ean
op
e
rati
on
rem
ov
es
ins
ulate
d
pi
xels
(ind
i
vidu
al
1s
that are surr
ound
e
d
by
0s
)
, s
uc
h
as the cen
te
r
pix
el
w
it
h
val
ue
1s
bo
a
r
ded
by zero
s
. In
ad
diti
on
, t
hin
ope
rati
on
rem
ov
es
pix
el
s
so
that
a
n
ob
j
e
ct
with
ou
t
hole
s
shrin
ks
to
a
m
ini
m
ally
connecte
d
ra
p,
an
d
an
ob
j
ect
with
hole
s
sh
ri
nk
s
to
a
c
onnected
rin
g
half
way
betwe
en
each
hole
and
the
oute
r
boun
dar
y.
I
ndee
d,
e
r
os
io
n
ope
rati
on
aff
ect
s
a
bin
a
r
y.
Im
age
by
erode
a
way
th
e
bounda
ries
of
fore
groun
d
pix
el
s
(
usual
ly
the
wh
it
e
pi
xels).
Th
us
, a
reas
of
foregr
ound
pixe
ls sh
rin
k
in
size, an
d
"
hole
s"
w
it
hi
n t
hose a
reas
beco
m
e large
r [15
,
16].
6.
PROP
OSE
D WOR
K
Mor
phologica
l
i
m
age
in
this
sect
ion
,
we
propose
(F
RB
S
)
sta
nd
f
or
Foot
pr
i
nt
Re
co
gn
it
ion
Bi
om
et
ric
Syst
e
m
in
wh
i
ch
it
is the ai
m
o
f
t
his p
a
per.
We enf
or
ce t
he
syst
e
m
an
d
de
velo
p
the
perf
orm
ance o
f
the
r
el
at
ed
work
by
m
erg
ing
betwee
n
im
age
proce
ssin
g
and
An
t
Col
ony
Op
ti
m
iz
ation
Algo
rithm
(
ACO
).
A
nt
Co
lon
y
Op
ti
m
iz
ation
Algorithm
(AC
O)
sel
ect
s
th
e
op
ti
m
al
feat
ur
e
,
w
hich
a
ffec
ts
the
res
ults
of
t
his
ty
pe
of
syst
em
.
As well
as,
t
o
c
reate Vis
ual F
ootp
rint
Databas
e.
T
he pr
opos
e
d work i
s s
how
n
in
F
ig
ure
4
.
Figure
4
.
Pro
pose
d
w
ork
f
l
ow
digram
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4
0
6
0
-
4
0
6
8
4064
6.1.
Visua
l
dat
aba
se a
c
quisi
tio
n
Footp
rint
dataset
is
rar
el
y
a
va
il
able
due
to
di
ff
ic
ulty
to
get
cl
ear
f
oo
t
pr
i
nt
i
m
ages
for
var
i
ou
s
usa
ge
.
This
is
sta
te
d
as
the
m
a
in
chall
eng
e
f
or
foo
tprin
t
rec
ogniti
on
researc
h.
D
ue
to
the
diff
i
culty
of
obta
in
ing
a
database
of
re
ady
-
m
ade
universal
feet.
Li
ve
im
ages
of
the
le
ft
F
oot
i
m
age
we
re
colle
ct
ed
f
r
om
nin
e
ind
ivi
du
al
s
.
Te
n
im
age
per
pe
rson
was
ta
ke
n
with
di
ff
e
ren
t
directi
ons.
T
he
se
i
m
ages
we
re
cl
assifi
ed
as
nine
cl
asses n
am
ed
foots. Eac
h wit
h
10 im
a
ges
w
e take these im
ages fr
om
m
ale
and
fem
al
e,
with an
a
ge va
ry f
r
om
25
to
60
ye
ar
s.
The
capt
ur
e
d
i
m
ages
are
hi
gh
res
olu
ti
on
a
nd
ta
ken
by
can
on
110
scan
ne
r
of
30
0
pix
el
qual
it
y
and
black
bac
kgr
ound.
T
he
foot
i
m
ages
w
ere
save
d
with
(1436
x3196)
pix
el
siz
e
an
d
JPG
f
orm
at
.
E
ach
te
n
i
m
age
for
per
s
on
are
cl
assi
fied
as
cl
us
te
rs
nam
ed
Foot.
F
ro
m
Fo
ot
1
to
Foot1
0
a
sam
ple
from
this
dataset
is
sh
ow
n
in
F
i
gur
e
5
.
Figure
5
.
A sa
m
ple o
f
f
oo
t i
m
age f
r
om
d
at
abase
6.2.
Visua
l
dat
aba
se pr
ocessin
g
At
first
RGB
i
m
age
is
conve
rted
to
t
he
gra
y
fr
om
then
to
the
bin
a
ry
f
orm
,
then
im
age
is
resized
to
(18
0x180)
pixe
l.
Im
age
is
de
noise
d
a
nd
cl
eared
t
he
n
it
is
bin
a
rized
with
Ca
nn
y
filt
er
to
detect
the
f
ootp
rint
edg
e
s.
I
n
ord
er
to
dete
rm
i
ne
the
f
oot
ti
p
the
re
su
lt
an
t
i
m
age
is
entered
to
m
or
phol
og
y
op
e
rat
ion
s
.
These
operati
on
s
are
cl
os
e,
thin,
e
r
os
io
n
and
cl
ean.
The
se
gm
entation
is
ob
ta
i
ne
d
on
footp
rint
i
m
age
represe
nted
by
three
blo
c
ks
the
fir
st
for
fi
nger
,
the
sec
ond
bl
ock
f
or
ce
nter
wh
il
e
the
third
is
for
the
heel
.
The
fir
st
and
t
he
thir
d
bl
ock
s
are
us
e
d
in
th
e
nex
t
sta
ge
for
rec
ogniti
on.
Wh
il
e,
we
ig
nore
the
sec
ond
bloc
k
because
it
d
i
d no
t
giv
e
acc
ur
a
te
an
d cl
ear i
nfor
m
at
ion
for re
cogniti
on
6.3.
Fo
ot
prin
t
feat
ure ex
tr
act
io
n
stag
e
Footp
rint
feat
ur
e
ext
racti
on
is
the
im
po
rtant
sta
ge
in
a
ny
bi
om
et
ric
syst
e
m
.
W
e
ha
ve
us
e
t
he
geo
m
et
ry
featu
res
of
t
he
foot
ti
p
via
reg
i
on
pro
per
ty
.
We
c
hoos
e
the
m
os
t
accu
rate
feat
ures
by
analy
sis
an
d
exp
e
rim
ents
as
sho
wn
bel
ow.
CEN
TROI
D
[
15,
17
]
determ
in
es
the
ce
nter
of
the
wh
it
e
pi
xel
re
gion.
It
r
esults
into
tw
o
co
ord
inate
.
The
first
is
the
horiz
onta
l
coo
r
din
at
e
(or
x
-
c
oord
i
nate)
of
the
ce
nte
r,
wh
il
e
the
se
co
nd
el
e
m
ent
is
the
ver
ti
cal
coor
din
at
e
(or
y
-
co
or
din
at
e)
.
ARE
A
ref
ers
t
o
the
a
ct
ual
nu
m
ber
of
pix
e
ls
re
pr
e
sents
the
re
gi
on.
O
RIENT
AT
ION
[15,
16]
refe
r
s
to
the
a
ngle
(in
de
gr
ees
a
lt
ern
at
ing
fro
m
-
90
to
90
de
gr
ee
s
)
a
m
on
gst
the
x
-
axis
a
nd
the
m
ajor
a
xis
of
the
re
gion.
PERI
MET
ER
com
pu
te
s
the
dista
nc
e
ar
ound
the
bo
a
r
der
of
t
he
re
gion
[17].
T
he
dist
ance
bet
w
ee
n
each
ad
joini
ng
pair
of
pix
el
s
arou
nd
t
he
bor
der
of
the
r
egio
n
com
pu
te
s
the
per
im
et
er.
In
this,
syst
em
five
featur
es
a
re
extracte
d
relat
ed
to
the
fi
ng
e
rs
segm
ent
an
d
fiv
e
featur
e
s
for
t
he
heel.
T
he
tot
al
is
te
n
featu
r
es
pe
r
im
age
wh
il
e
we
ha
ve
te
n
im
age
pe
r
pe
rson,
s
o
we
will
captu
re
one
hundre
d
feature
to
each
pe
rson
of
the
ni
ne
vo
lun
te
ers
.
Furth
er,
900
feat
ur
e
s
are
colle
ct
ed
and
save
d
in
E
XC
EL
file
were
the
featu
re
da
ta
base
is
co
nst
ru
ct
ed
t
o
be
us
e
la
te
r
at
the
te
st
ste
p
duri
ng
recog
niti
on
sta
ge.
A
s
am
ple f
ro
m
f
eat
ur
es
dat
abase is s
ho
wn in T
able
2.
6.4.
Recogni
tion
te
chniques
st
age
We
us
e
te
n
im
age
to
eve
ry
pe
rson
f
or
t
he
te
st
pr
oce
ss
w
he
re
it
s
featur
es
are
save
d
in
the
data
base
featur
e
s.
T
he
la
st
i
m
age
of
th
e
pe
rson
is
us
e
d
in
the
syst
em
,
as
a
qu
e
ry
i
m
age
we
re
it
s
featur
e
s
ar
e
e
xt
racted
on
li
ne
duri
ng
t
he
syst
em
op
erati
on.
T
he
qu
ery
featu
res
ar
e
com
par
ed
with
al
l
the
data
base
fe
at
ur
e
s
withi
n
An
t
Col
on
y
a
lgorit
hm
.
Fr
om
a
giv
en
im
age,
the
accurate
featu
res
ar
e
extracte
d
and
are
el
im
inate
d.
This
li
m
it
ation
is
d
e
rive
d
thr
ough
m
inu
te
or
no
a
na
ly
ti
cal
data,
an
d
dis
reg
a
r
d
the
fire
d
f
eat
ur
es
.
ACO
cal
c
ulate
the
fitness
fun
ct
ion
with
t
he
m
ini
m
u
m
diff
eren
ce
(m
in)
be
tween
qu
e
ry
a
nd
databa
se
fe
at
ur
es
F
igure
6
s
how
the m
ai
n
proce
dure
of A
C
O p
rogr
am
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A h
y
pr
id
te
ch
ni
qu
e f
or
hum
an
foo
t
pr
int rec
ogniti
on (Y
ah
y
a Is
ma
il
Ib
r
ahim)
4065
Table
2.
A
sam
ple from
EXC
EL f
il
e that
re
presents
f
eat
ure
s d
at
a
base
Figure
6
.
Mai
n al
gorithm
o
f ACO
pro
gram
A
segm
ent
of
our
pro
posed
work
pro
gr
am
is
giv
en
in
F
ig
ur
e
7
wer
e
the
input
is
the
ACO
par
am
et
er
and
the
outp
ut
is
the
m
at
ched
i
m
age
with
it
is
fr
e
quency
w
it
hin
the
datab
ase
ot
herwise
a
m
essage
of
r
ejects.
Excel
file
newfoot
co
ntain
i
m
age
footp
rint
featu
res
wh
ic
h
e
ntered
to
A
CO
f
unct
ion.
T
his
f
unct
ion
ret
urns
t
he
b
est
r
oute
an
d
the
best
in
de
x
wh
ic
h
c
om
par
ed
with
10
0
that
ad
dress
t
he
e
ntit
ie
s
within
a
co
rr
e
spo
nd
i
ng
cl
us
te
r foot
de
no
te
d from
1
to 10.
pr
op
os
e
d
w
or
k pr
ogram
Inp
ut
:
AC
O’
s
pa
r
amet
ers (f
eatures)
Ou
tp
ut
:
M
atch
ed
fo
otpri
nt im
ag
e
wi
th it
s in
de
x wi
thin the
data
base
[
per
l,
hea
der
]
=xlrea
d(
'
newf
oot'
)
[b
sti
nd
x,
gbest
]
=aco(p
e
rl,
dat
a)
Bst
ind
x;
gbest
; i
4=10
Tt(I,1)=
gbest;
Tt(I,2)=
bs
ti
nd
x;
Tt(I,3)=ceil
(
bst
ind
x/i4
);
If
(
gb
est
<
100)
X=
bs
ti
nd
x/i4;
Seq=
cei
l(
bs
ti
ndx/i4
);
Else
Seq=
0;
En
d
If
s
eq
!=0
Ms
gbox
of
ac
cepta
nce wit
h
t
he
im
age
num
be
r
Else
Ms
gbox
o
f
re
je
ct
wi
th the image
nu
mb
e
r
end
Figure
7
.
Se
gme
nt of t
he pr
opos
e
d
w
ork
pr
ogram
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4
0
6
0
-
4
0
6
8
4066
7.
E
X
PERI
MEN
TAL O
UTC
O
MES
The
ef
fici
ency
and
t
he
e
nh
a
ncem
ent
of
ou
r
pro
pose
d
w
ork
a
re
e
valuat
ed,
by
m
any
e
xp
e
rim
ents.
These
e
xperim
ents w
ere writ
te
n
a
nd
r
un o
n M
at
la
b2
01
7a.
Th
e recog
niti
on r
esults of
the
query
im
age
with
th
e
visu
al
databas
e
are
sho
wn
i
n
Fig
ure
8
.
A
le
tt
er
is
sh
ow
n
in
a
m
essag
e
box
if
rec
ogniti
on
resu
lt
s
i
s
true
or o
t
herwise fa
lse
.
Figure
8
.
The
re
su
lt
of
qu
e
ry
with
visu
al
dat
abase im
age
We
qu
a
ntit
y
the
qu
al
it
y
of
the
ou
tc
om
es
via
two
process
es,
nam
el
y,
recall
and
pr
e
ci
sion
t
hey
are
de
fin
e
d
as
(
1) an
d (2)
r
especti
vely
[18]
.
Re
cal
l = TR
(i)/M
(
i)
(1)
Pr
eci
sio
n=TR(
i)/TR
(i)+FF
(i)
(
2)
Her
e
,
M(i)
re
f
ers
to
the
f
requen
cy
of
im
ages
in
t
he
cl
ass,
TR(i)
is
the
num
ber
of
im
ages
fitt
ing
ly
retrieve
d,
FF(
i
)
is
the
nu
m
be
r
of
im
ages
m
ist
aken
ly
retrie
ved.
Ta
ble
3
de
scribe
t
he
qu
al
it
y
m
easur
es
for
a
sam
ple
of
im
a
ges.
Table
1
de
scribe
that
A
CO
sel
ect
s
th
e
best
featu
res
wh
il
e
obta
ined
the
qual
it
y
m
easure
resu
lt
s
as
s
how
n
in
F
ig
ure
6
.
Wh
e
re
t
he
ei
ght
fir
st
cl
us
te
re
s
gi
ve
1
00%
a
ccuaracy
f
or
re
cal
l
and
preci
sion
bu
t
90%
f
or
recall
and
100%
f
or
pr
eci
sio
n
for
the
la
st
cl
us
te
r
.
Be
cause,
only
on
e
im
age
is
no
t
rele
va
nt
in
that
cl
ass. T
he
a
verage is
98.
88
%
and 10
0% for
r
ecal
l and p
recisi
on over
all
v
i
su
al
database
r
ecspecti
vely
.
Table
3.
T
he
quali
ty
m
easur
e
s
Clas
s ID
Recall
Precisio
n
Fo
o
t 1
100%
100%
Fo
o
t 2
100%
100%
Fo
o
t 3
100%
100%
Fo
o
t 4
100%
100%
Fo
o
t 5
100%
100%
Fo
o
t 6
100%
100%
Fo
o
t 7
100%
100%
Fo
o
t 8
100%
100%
Fo
o
t 9
90%
100%
Av
erage
9
8
.88
%
100%
Wh
e
re
t
he
ei
ght
fir
st
cl
asses
giv
e
10
0%
ac
cuar
acy
for
re
cal
l
and
preci
s
ion
but
90%
f
or
recall
an
d
100%
for
preci
sion
for
the
la
s
t
cl
ass.
The
a
ve
rag
e
is
98.88
%
and
100%
f
or
recall
and
pr
eci
sion
rec
sp
ec
ti
vely
.
Be
cause,
only
on
e
im
age
is
not
rele
van
t
i
n
t
hat
cl
ass.
The
pro
ps
ed
w
ork
is
com
par
ie
d
with
oth
e
r
wor
ks
on
footp
rint
rec
og
niti
on
t
his
com
par
is
on
is
disc
ribe
d
in
Ta
ble
3.
T
hese
w
ork
s
wer
e
bas
ed
on
di
ff
e
ren
t
m
eth
ods
bu
t
not
feat
ur
e
sel
ect
ion
with
A
NT
C
OL
O
NY
O
ptim
iz
at
i
on
al
gorithm
.
The
rec
ogniti
on
rate
s
how
t
ha
t
our
work
gai
n
a
superi
or
resu
lt
s
a
ppr
ox
im
at
ed
to
98.88%
a
nd
100%
.
T
hese
re
su
lt
s
show
t
he
ro
le
of
op
ti
m
aizati
on
(
AC
O)
feat
ur
e
sel
ect
ion
wh
i
ch
sel
ect
s
the
best
featu
res
in
a
s
m
al
l
feat
ur
e
acc
ur
at
e
s
et
.
W
e
c
om
par
e
these
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
A h
y
pr
id
te
ch
ni
qu
e f
or
hum
an
foo
t
pr
int rec
ogniti
on (Y
ah
y
a Is
ma
il
Ib
r
ahim)
4067
resu
lt
s
with
r
esearch
dep
e
nd
on
geo
m
et
ry
featu
res
te
c
hn
i
qu
e
s
how
n
in
T
able
4.
They
are
Foot
Sh
ape
Geo
m
et
ry
and
ANN
[25]
an
d
Sh
a
pe
descr
i
pt
or
s
[16]
wit
h
recog
niti
on
rate
92.
5%
a
nd
75.8%
res
pecti
vely
.
Our
pro
posed
w
ork
s
how
bes
t
recog
nation
r
at
e
than
them
bo
t
h.
A
ls
o,
ta
bl
e
4
s
how
that our work
is
t
he
best
i
n
com
par
ison wi
th v
a
rio
us
kind
of t
ech
niques
w
e
re th
ei
r
r
es
ults ra
ng
e
d bet
ween 5
1%
to 9
2.8.
Ta
ble
4.
Per
for
m
ance ev
al
uation bet
ween t
he
Propose
d Wor
k
a
nd anothe
r r
esearch
Au
th
o
r
&
Y
ear
Techn
iq
u
e
Reco
g
n
itio
n
Rate %
Jan
g
J.
R. 19
9
3
[
1
9
]
Co
m
p
r
eh
en
siv
e E
v
alu
atio
n
M
o
d
el
9
2
.8%
Nak
aji
m
a
,
2
0
0
0
[
2
0
]
Euclid
ean d
istan
ce
85%
A.
Kad
y
arov
an
d
M
.
Petrou
s 2
0
0
1
[
2
1
]
Tr
ace
T
rans
f
o
r
m
T
echn
iq
u
e
Distin
g
u
ish
sp
ecie
s o
f
ins
ects to
a
sp
ecif
ied
deg
ree
Jin
-
W
o
o
et.
Al.
,20
0
3
[
2
2
]
Hid
d
en
M
arko
v
M
o
d
el
64%
Jaes
o
ek
Yun
et.
Al
.
2
0
0
7
[
2
3
]
Neu
ral
N
etwo
rk
92%
B.S.Ch
in
,et.
Al.
2
0
0
7
[
2
4
]
ART2
Algo
rith
m
Fair
p
erfo
r
m
an
ce i
n
clus
tering
Ku
m
ar
V.D.
,and
M.
Ra
m
ak
rish
n
an
,
2010.
[
1
]
DCT
FT
SHT
8
3
.64
%
8
7
.43
%
9
2
.37
5
%
Jacq
u
elin
e Do
m
ja
n
ic
2013
[
1
0
]
The in
tra
cl
ass
cor
relation
coef
f
icien
ts
(I
CC
)
f
o
r
th
e f
irst
f
o
u
r
p
rincip
al co
m
p
o
n
en
ts
95%
85%
8
5
%, and
88
%
Hash
e
m
Kadh
i
m
m
.
an
d
Ghali Fatia
m
a
2
0
1
6
[
2
5
]
Fo
o
t Shap
e Geo
m
e
try
an
d
ANN
9
2
.5%
Riti Ku
sh
wah
a,
et.
Al.
2
0
1
7
[
1
6
]
Sh
ap
e des
cripto
rs
Sh
ap
e seg
m
en
tatio
n
Eigen
f
o
o
t
75
.
8
%.
5
1
%
9
2
.%%
The Prop
o
sed
W
o
r
k
Geo
m
etr
y
Featu
res
100%
8.
CONCL
US
I
O
N
We
pr
opos
e
d
a
hype
r
al
gori
thm
wh
ic
h
m
e
rg
e
betwee
n
i
m
age
processi
ng
te
c
hniq
ues
for
feat
ur
e
extracti
on
an
d
ACO
for
feat
ur
e
sel
ect
ing
ba
sed
he
ur
ist
ic
strat
egy.
Im
age
pr
oce
ssin
g
r
eflect
s
best
featur
es
extracti
on w
it
h
bin
a
rizat
ion
a
nd
m
or
phol
ogy
op
e
rati
ons.
ACO
sho
w
an
ex
cel
le
nt
feat
ures
in
te
rm
s
of
sm
a
ll
es
t
featur
e
s
nu
m
ber
.
T
he
ex
per
i
m
ental
ou
tc
om
es
on
im
age
data
set
s
sho
w
the
s
up
e
rio
r
it
y
of
this
pro
cess
i
n
com
par
esi
on wi
th anothe
r wor
k on the
sam
e fiel
d.
T
he
r
eco
gnit
ion
rate i
s e
qu
al
t
o 98.8%
and 10
0%.
REFER
E
NCE
S
[1]
Kum
ar
V.
D.
a
nd
Ramakrishna
n
M.,
“
Footprint
Rec
ogn
it
ion
usi
ng
Modifie
d
Se
quent
i
al
Ha
ar
E
ner
g
y
Tr
ansform
(MS
HET)
,”
I
JC
SI
Inte
rnat
ional
Journal
of
Computer
Sc
ie
nc
e
Iss
u
es
,
v
ol
/i
ss
ue:
7
(3
)
,
2010
.
[2]
Kus
hwaha
R.
,
e
t
al
.
,
“
Deta
i
le
d
Anal
y
sis
of
Footprint
Geom
e
tr
y
for
Person
Id
ent
ifica
ti
on
,”
13
th
Inte
rnat
ional
Confe
renc
e
on
S
ignal
-
Image
Te
c
hnology
&
Inte
rnet
-
Based
Syst
e
ms
(
SI
TIS)
,
2017.
[3]
I
.
M.
Alsaa
di,
“
Phy
siologi
ca
l
Biom
et
ric
Authent
i
cation
S
y
stems
,
Advanta
ges,
Disadva
ntages
and
Futur
e
Deve
lopment : A
Revi
e
w
,
”
In
ter
nati
onal Journal
of
Scientifi
c
&
Technol
ogy
R
ese
arch,
vol
/i
ss
ue:
1
(
1
)
,
2008
.
[4]
Al
-
Dulai
m
i
K
.
A.,
“
Us
ing
Fea
ture
Ext
ra
ction
for
Hum
an
Footprint
s
Re
cog
nit
ion,
”
In
te
rnat
ional
Journal
o
f
Computer
Appli
cat
ions,
v
ol
/
issue:
64
(3)
,
2013
.
[5]
Kum
ar
K.
N
.
an
d
Dube
y
S
.
,
“
Bi
om
et
ric
Auth
entication
using
Hum
an
Footprint
,
”
Foundat
ion
of
C
omputer
Scienc
e
FCS,
N
ew Y
ork,
USA,
v
ol
/i
ss
ue:
3
(7),
2012.
[6]
Oppus
C.
,
“
Im
a
ge
edge
de
tecti
o
n
using
ant
col
o
n
y
opt
imiza
t
ion,
”
We
st
Scientifi
c
and
Engi
ne
eri
ng
Ac
ademy
an
d
Soci
e
ty
(
WS
EA
S
)
Tr
an
sacti
ons o
n
Signal P
roc
essing
,
2010
.
[7]
T
.
Kavita,
“
Ant
col
on
y
op
ti
m
iza
ti
on
al
gori
thm:
adva
nt
ag
es,
app
l
ic
a
ti
ons
and
challe
ng
es
,”
Instit
u
t
e
of
Technol
ogy
and
Manage
men
t
Univ
erse,
Vado
dara Gujarat.
C
omputer
Mode
l
ling
&
New
Techn
ologi
es
,
v
ol
/
issue:
21
(
2
)
,
2017
.
[8]
Kum
ar
K.
N
.
and
Dube
y
S.
,
“
Mathe
m
atical
m
odel
ing
of
footpr
in
t
base
d
biometric
Rec
ogni
ti
on
,
”
Inte
rnational
o
f
mathemati
ca
l
tre
nds and
technol
ogy
(
IJM
IT)
,
v
ol
/i
ss
ue:
54
(
6
)
,
20
18.
[9]
Uhl
A.
and
W
il
d
P
.
,
“
Footprint
-
b
ase
d
biometr
ic v
eri
fi
ca
t
ion,
”
Jou
rnal
of El
e
ct
roni
c
Imaging
,
2008
.
[10]
J
.
Dom
ja
nic
,
e
t
al
.
,
“
Geom
et
ric
m
orphom
et
ric
footpri
nt
an
aly
sis
of
y
oung
wom
en,
”
Journal
of
Foot
and
Ankle
Re
search
,
vol
.
6
,
pp.
27
,
2013
.
[11]
Katiy
ar
S.
,
e
t
a
l
.
,
“
Ant
Colon
y
Optimiza
ti
o
n:
A
Tut
ori
al
Rev
iew
,”
MR
In
te
rna
ti
onal
Journ
al
o
f
Engi
n
ee
ring
&
Technol
ogy
,
v
ol
/
issue:
7
(
2
)
,
2011
.
[12]
M
.
Dorigo
and
T
.
Stutzle,
“
Ant
Colon
y
Optimi
za
t
ion,
”
The
MI
T
Press,
Londo
n,
Engl
and,
200
4
Mass
achuse
tt
s
Instit
ute of
Tech
nology
,
2004
.
[13]
Manie
z
zo
V.
,
et
al.,
“
Ant
Colo
n
y
Op
ti
m
iz
a
ti
on
.
In:
New
Opti
m
iz
at
ion
Techni
ques
in
Engi
n
eeri
ng
,”
Studies
i
n
Fuzziness and
S
oft
Computing
,
S
pringe
r, Ber
li
n,
Heide
lb
erg
,
v
ol
.
141
,
2004
.
[14]
A
.
S
.
Abdulja
b
ar,
“
Us
ing
a
N
um
ber
of
Modifie
d
CG
Al
gori
t
hm
s
in
Patt
ern
Rec
ognition
,
”
P
h.
D.,
The
sis
in
Com
puta
ti
onal
Mathe
m
at
i
cs,
C
oll
ag
e
of
Mathem
at
ic
a
l and
Co
m
pute
r
Scie
n
ce
s
,
Univer
si
t
y
of
Mos
ul,
2018.
[15]
R
.
C.
Gonza
lez
a
nd
R
.
E. Woods
,
“
Digit
al i
m
age
proc
essing
,”
2
nd
Ed.
,
pre
n
ti
c
e
–
H
al
l
,
New Jers
e
y
,
US
A,
2002.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
4
0
6
0
-
4
0
6
8
4068
[16]
R
.
Kus
hwaha
,
et
al
.
,
“
Person
identifi
c
ation
on
the
basis
of
footpri
nt
geometr
y
,
”
12th
Inte
rnationa
l
Confe
renc
e
on
Signal
-
Image
Te
chnol
ogy
&
Inte
rnet
-
Based
S
ystem
s
,
2016
.
[17]
Nick
Eff
ord
,
“
Digit
al
Im
age
Proce
ss
ing:
A
Pr
ac
tical
In
trod
uct
ion
Us
ing
J
ava
,”
P
earson
Educ
ati
on
,
200
0.
Avail
ab
le
:
htt
ps:
//
ww
w.c
s.a
u
ckl
a
nd.
ac.nz/course
s
/c
om
psci773s1c/lecture
s/Im
age Proce
ss
ing
-
html/
[18]
H
.
Mulle
r
,
e
t
al
.
,
“
Perform
anc
e
E
val
ua
ti
on
in
CBI
R:
Overvi
ew
an
d
Propos
al
s
,”
A
te
chn
ic
al
report
subm
it
te
d
to
the
Computing
Sc
ience
C
ent
er
,
Uni
v
ersity
o
f
Gen
ev
a
,
Gene
va
,
Sw
it
zer
land
,
1999
.
[19]
J.
R.
Jang,
“
ANFIS:
Adapti
ve
-
n
et
work
-
base
d
fu
zzy
infe
r
enc
e
s
ystem
,”
I
EE
E
Tr
ans.
Syst
.
,
v
ol.
23,
1993.
[20]
K.
Naka
ji
m
a,
et
al
.
,
“
Fo
ot
-
B
ase
d
Personal
Rec
ogn
it
io
n
,
”
IEEE:
Tr
.
On
Bi
omedi
cal
Engi
n
ee
rin
g
,
v
ol
/i
ss
ue:
47
(
11
)
,
2000
.
[21]
A.
Kad
y
rov
and
M.
Petrou,
“
The
Tra
c
e
Tr
ansform
and
Its
Appl
ic
a
ti
ons,”
I
EE
E
Tr
ansacti
ons
on
Pat
te
rn
Analysi
s
and
Mac
hin
e
In
t
el
li
g
ence
,
v
ol
/
issue:
23
(
8
)
,
pp
.
81
1
-
828,
2001
.
[22]
W
.
Jungl,
et a
l
.
,
“
Dy
n
amic
-
Footp
rint
b
ase
d
Perso
n
Ide
nt
ifi
c
ation
using Ma
t
-
t
y
p
e P
ressure
Sensor
,
”
IE
EE
,
2003
.
[23]
V.
A.
Kum
ar
a
nd
M.
Ramakris
hnan,
“
L
ega
c
y
of
F
ootpri
nts
R
ec
ogni
ti
on
-
A
R
evi
ew
,
”
In
te
rnat
ional
Journal
o
f
Computer
Appli
cat
ions,
v
o
l
/
issue:
35
(
11
)
,
pp
.
9
-
16,
2011
.
[24]
B.
S.
Shin
,
e
t
a
l.
,
“
Segm
ent
at
ion
of
Scanne
d
Inse
ct
Footprin
ts
Us
ing
ART2
for
T
hre
shold
Select
i
on,
”
LN
CS
4872
,
Springer
-
Verl
ag
,
2007.
[25]
H
.
Kadhim
M
.
and
G
.
Fati
ama,
“
Hum
an
Ide
nti
ficat
ion
Us
ing
Hum
an
Footprint
Feat
ur
es
,
”
Engi
nee
ring
a
nd
Manufac
turing
,
v
ol.
4
,
2016
.
Evaluation Warning : The document was created with Spire.PDF for Python.