I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
9
,
No
.
2
,
A
p
r
il
201
9
,
p
p
.
1
0
2
1
~
1
0
2
7
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v9
i
2
.
pp
1
0
2
1
-
1027
1021
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
Propo
sed
a
lg
o
rith
m
for i
m
a
g
e clas
s
ificatio
n using
regres
sio
n
-
ba
sed
pre
-
pro
cess
ing
an
d recog
nition
m
o
dels
Cha
nin
t
o
rn
J
it
t
a
w
iriy
a
nu
k
o
o
n
A
s
su
m
p
ti
o
n
Un
iv
e
rsity
,
T
h
a
il
a
n
d
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
A
p
r
2
7
,
2
0
1
8
R
ev
i
s
ed
Sep
3
0
,
2
0
1
8
A
cc
ep
ted
O
c
t
15
,
2
0
1
8
Im
a
g
e
c
las
si
f
ic
a
ti
o
n
a
lg
o
rit
h
m
s
c
a
n
c
a
te
g
o
rize
p
ix
e
ls
r
e
g
a
rd
in
g
ima
g
e
a
tt
rib
u
tes
w
it
h
th
e
p
re
-
p
ro
c
e
ss
in
g
o
f
le
a
rn
e
r’s
train
e
d
sa
m
p
les
.
T
h
e
p
re
c
isio
n
a
n
d
c
las
sif
ic
a
ti
o
n
a
c
c
u
ra
c
y
a
re
c
o
m
p
lex
to
c
o
m
p
u
te
d
u
e
to
th
e
v
a
riab
le
siz
e
o
f
p
ix
e
ls
(d
if
fe
re
n
t
im
a
g
e
w
id
th
a
n
d
h
e
ig
h
t)
a
n
d
n
u
m
e
ro
u
s
c
h
a
ra
c
teristics
o
f
im
a
g
e
p
e
r
se
.
T
h
is
re
se
a
r
c
h
p
ro
p
o
se
s
a
n
i
m
a
g
e
c
las
si
f
ic
a
ti
o
n
a
lg
o
rit
h
m
b
a
se
d
o
n
re
g
re
ss
io
n
-
b
a
se
d
p
re
-
p
r
o
c
e
s
sin
g
a
n
d
t
h
e
re
c
o
g
n
it
i
o
n
m
o
d
e
ls.
T
h
e
p
ro
p
o
se
d
a
lg
o
rit
h
m
f
o
c
u
se
s
o
n
a
n
o
p
t
im
iza
ti
o
n
o
f
p
re
-
p
ro
c
e
ss
in
g
re
su
lt
s
su
c
h
a
s
a
c
c
u
ra
c
y
a
n
d
p
re
c
isio
n
.
T
o
e
v
a
lu
a
te
a
n
d
v
a
li
d
a
te,
th
e
re
c
o
g
n
it
i
o
n
m
o
d
e
l
is
m
a
p
p
e
d
i
n
o
rd
e
r
t
o
c
l
u
ste
r
t
h
e
d
ig
it
a
l
im
a
g
e
s
w
h
ich
a
re
d
e
v
e
lo
p
i
n
g
th
e
p
ro
b
lem
o
f
a
m
u
lt
id
im
e
n
sio
n
a
l
sta
te
sp
a
c
e
.
S
i
m
u
latio
n
re
su
lt
s
sh
o
w
th
a
t
c
o
m
p
a
re
d
to
e
x
isti
n
g
a
l
g
o
rit
h
m
s,
th
e
p
ro
p
o
se
d
m
e
th
o
d
o
u
tp
e
rf
o
r
m
s
w
it
h
th
e
o
p
ti
m
a
l
a
m
o
u
n
t
o
f
p
re
c
isio
n
a
n
d
a
c
c
u
ra
c
y
in
c
las
si
f
ica
ti
o
n
a
s
we
ll
a
s
re
su
lt
s
h
ig
h
e
r
m
a
tch
in
g
p
e
rc
e
n
tag
e
b
a
se
d
u
p
o
n
im
a
g
e
a
n
a
l
y
ti
c
s.
K
ey
w
o
r
d
s
:
C
las
s
i
f
icatio
n
I
m
ag
e
e
m
b
ed
d
er
R
ec
o
g
n
itio
n
m
o
d
el
R
eg
r
es
s
io
n
-
b
ased
ap
p
r
o
ac
h
Si
m
u
latio
n
Co
p
y
rig
h
t
©
2
0
1
9
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
C
h
a
n
i
n
to
r
n
J
itta
w
ir
i
y
an
u
k
o
o
n
,
Ass
u
m
p
t
io
n
U
n
iv
er
s
it
y
,
Sa
m
u
t P
r
ak
an
,
T
h
ailan
d
.
E
m
ail: c
h
an
i
n
to
r
n
j
tt
@
au
.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
An
i
m
a
g
e
d
o
es
n
o
t
co
n
ta
in
a
s
tr
u
ct
u
r
ed
f
o
r
m
o
f
d
ig
ital
d
a
ta
b
u
t
r
ath
er
a
n
u
n
s
tr
u
ct
u
r
ed
p
atter
n
.
I
n
o
r
d
er
to
an
aly
z
e
a
n
i
m
ag
e
a
n
d
cu
r
ate
s
u
ch
d
ata
h
o
w
ev
er
t
h
e
u
n
s
tr
u
ct
u
r
ed
f
o
r
m
o
f
d
ata
n
ee
d
s
to
b
e
co
n
v
er
ted
to
a
v
ec
to
r
r
e
p
r
esen
tatio
n
ac
co
r
d
in
g
l
y
.
Dee
p
lear
n
i
n
g
a
n
d
em
b
ed
d
in
g
m
et
h
o
d
s
h
elp
r
ec
o
v
er
th
e
n
u
m
er
i
c
p
atter
n
f
r
o
m
u
n
s
tr
u
ct
u
r
ed
d
ata.
T
h
e
q
u
alit
y
a
n
d
ac
cu
r
ac
y
o
f
clu
s
ter
i
n
g
i
m
ag
e
s
d
ep
en
d
o
n
f
ir
s
tl
y
t
h
e
ac
c
u
r
ac
y
of
p
r
e
-
p
r
o
ce
s
s
i
n
g
a
n
d
s
ec
o
n
d
l
y
t
h
e
s
u
b
s
eq
u
e
n
t
e
m
b
ed
d
in
g
a
lg
o
r
ith
m
[
1
]
.
A
n
i
m
a
g
e
i
s
s
eg
m
en
ted
in
o
r
d
er
to
class
i
f
y
t
h
e
p
ix
el
s
co
r
r
ec
tl
y
i
n
a
d
ec
is
io
n
-
m
a
k
i
n
g
ap
p
licati
o
n
.
T
h
is
ap
p
r
o
ac
h
is
b
en
e
f
ici
al
in
t
h
e
f
ield
o
f
th
e
co
m
p
lica
ted
i
m
a
g
e
s
u
ch
as
p
atter
n
r
ec
o
g
n
itio
n
,
m
ed
ical
i
m
ag
e
p
r
o
ce
s
s
i
n
g
,
o
r
tr
af
f
i
c
i
m
ag
e.
Di
f
f
er
en
t
ap
p
r
o
ac
h
es
f
o
r
i
m
a
g
e
s
e
g
m
e
n
tatio
n
d
escr
ib
ed
in
[
1
]
o
p
t
clu
s
ter
i
n
g
,
tr
ain
i
n
g
,
t
h
r
esh
o
ld
-
b
ased
v
alu
e
f
o
r
t
h
e
co
m
p
u
tatio
n
.
C
l
u
s
ter
i
n
g
tech
n
iq
u
es i
n
clu
d
e
f
u
zz
y
c
-
m
ea
n
s
(
FC
M)
an
d
K
-
m
ea
n
s
m
et
h
o
d
.
Far
h
an
g
[
2
]
h
as
in
tr
o
d
u
ce
d
a
n
e
w
K
-
m
ea
n
s
c
lu
s
ter
in
g
alg
o
r
ith
m
a
s
t
h
e
i
m
p
le
m
e
n
tatio
n
o
f
K
-
m
ea
n
s
alg
o
r
ith
m
p
er
s
e
i
s
s
i
m
p
le.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
is
ap
p
l
ie
d
to
f
ac
e
i
m
a
g
e
m
i
n
in
g
.
E
x
p
er
i
m
e
n
tal
i
m
a
g
es
ar
e
ch
o
s
en
f
r
o
m
t
h
e
d
atab
ase
s
y
s
te
m
.
T
h
e
ex
p
er
i
m
e
n
tal
r
esu
lts
i
m
p
r
o
v
e
ac
cu
r
ac
y
r
ate,
r
ed
u
ce
p
r
o
ce
s
s
in
g
ti
m
e
d
u
e
to
a
r
ed
u
cib
le
n
u
m
b
er
o
f
iter
at
io
n
s
an
d
s
m
aller
cl
u
s
t
er
d
is
tan
ce
.
Ho
w
ev
er
,
i
m
a
g
e
class
if
ica
tio
n
an
d
r
ec
o
g
n
itio
n
m
o
d
el
ar
e
n
o
t c
ite
d
.
T
h
ese
in
ter
est
in
g
o
cc
u
r
r
en
c
es n
ee
d
ex
tr
a
i
n
v
e
s
ti
g
atio
n
.
Xin
a
n
d
Sag
a
n
[
3
]
h
av
e
p
r
o
p
o
s
ed
an
i
m
ag
e
cl
u
s
ter
in
g
alg
o
r
ith
m
u
s
i
n
g
m
u
ltip
le
a
g
e
n
t
s
f
o
r
o
p
tim
izin
g
cl
u
s
ter
ce
n
t
e
r
.
T
h
e
alg
o
r
ith
m
b
ased
o
n
f
u
zz
y
l
o
g
ic
m
ap
s
t
h
e
i
m
ag
e
c
lu
s
ter
s
as
i
n
telli
g
en
t
ag
e
n
t
m
o
v
i
n
g
in
th
e
s
tate
s
p
ac
e.
Si
m
u
latio
n
r
es
u
lt
s
s
u
p
p
o
r
t
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
co
m
p
ar
ed
t
o
ex
is
ti
n
g
m
et
h
o
d
s
ca
n
o
p
ti
m
ize
a
n
u
m
b
er
o
f
cl
u
s
ter
ce
n
ter
s
,
th
e
n
u
m
b
er
o
f
ca
teg
o
r
ies
a
n
d
th
e
c
lass
if
ica
tio
n
ac
cu
r
ac
y
(
C
A
)
is
h
ig
h
er
.
T
h
e
s
ize
o
f
b
ig
d
ata
s
t
ate
s
p
ac
e
is
r
ed
u
ce
d
b
y
o
p
ti
m
i
zin
g
th
e
co
llectio
n
o
f
ce
n
t
e
r
.
T
h
e
p
ap
er
s
o
m
e
h
o
w
n
ee
d
s
to
ad
d
itio
n
all
y
co
n
s
id
er
th
e
p
r
o
ce
s
s
in
g
ti
m
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
0
2
1
-
1027
1022
T
h
is
p
ap
er
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
f
o
r
cla
s
s
i
f
y
in
g
d
i
g
ital
i
m
a
g
es
b
ased
u
p
o
n
p
r
e
-
p
r
o
ce
s
s
an
d
r
ec
o
g
n
itio
n
m
o
d
el
s
.
Fir
s
t
l
y
,
i
n
th
i
s
r
e
g
ar
d
s
,
w
id
el
y
u
s
e
d
th
r
ee
i
m
a
g
e
cla
s
s
i
f
icat
io
n
s
ar
e
ex
a
m
i
n
ed
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
ch
ar
ac
ter
i
z
ed
b
y
r
e
g
r
ess
io
n
-
b
a
s
ed
p
r
e
-
p
r
o
ce
s
s
in
g
a
n
d
r
ec
o
g
n
itio
n
m
o
d
els
is
i
n
tr
o
d
u
ce
d
.
Seco
n
d
l
y
,
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
co
m
p
ar
ed
to
t
h
r
ee
t
r
ad
itio
n
al
ap
p
r
o
ac
h
es
i
s
e
v
alu
ated
u
s
i
n
g
t
h
e
class
i
f
icatio
n
ac
cu
r
ac
y
(
C
A
)
an
d
p
r
ec
is
io
n
m
etr
ics.
T
h
ir
d
l
y
,
s
e
v
er
al
al
g
o
r
ith
m
s
u
s
ed
f
o
r
r
ec
o
g
n
itio
n
m
o
d
el
an
d
th
eir
e
v
alu
at
io
n
ar
e
s
cr
u
ti
n
i
z
ed
.
L
a
s
tl
y
,
e
x
p
er
i
m
e
n
tal
r
e
s
u
lt
s
an
d
a
n
al
y
s
is
ar
e
d
is
c
u
s
s
ed
in
o
r
d
er
t
o
s
tate
th
e
f
u
r
t
h
er
r
esear
ch
in
v
e
s
ti
g
ati
o
n
.
2.
I
M
AG
E
CL
A
SS
I
F
I
CA
T
I
O
N
Sev
er
al
ap
p
r
o
ac
h
es
h
a
v
e
b
ee
n
p
r
esen
ted
to
i
m
p
r
o
v
e
th
e
cla
s
s
if
ica
tio
n
q
u
al
it
y
o
f
d
ig
ita
l
i
m
ag
es.
T
h
e
alg
o
r
ith
m
d
ev
elo
p
ed
b
y
f
u
zz
y
lo
g
ic
to
d
ea
l
w
i
th
a
m
b
ig
u
it
y
in
d
ig
i
tal
i
m
ag
e
s
[
4
]
,
[
5
]
.
T
h
e
class
i
f
icatio
n
an
d
s
p
ac
e
v
ec
to
r
r
elatio
n
s
h
ip
h
a
v
e
b
ee
n
in
s
p
ec
ted
b
ased
o
n
Ma
r
k
o
v
ia
n
m
o
d
e
ls
a
s
i
n
tr
o
d
u
ce
d
in
[
6
]
,
[
7
]
.
T
h
e
h
ier
ar
ch
ical
cla
s
s
i
f
icat
io
n
h
a
s
b
ee
n
also
e
m
p
lo
y
ed
f
o
r
i
m
ag
e
c
lass
if
ica
tio
n
.
A
r
ti
f
icia
l
I
n
telli
g
en
ce
(
A
I
)
tech
n
o
lo
g
y
h
a
s
b
ee
n
o
p
ted
to
ch
o
o
s
e
t
h
e
v
ar
iab
les
in
o
r
d
er
to
in
cr
ea
s
e
t
h
e
e
x
clu
s
i
v
e
cl
as
s
i
f
icatio
n
q
u
alit
y
.
T
h
e
n
eig
h
b
o
r
h
o
o
d
d
ec
is
io
n
is
in
tr
o
d
u
ce
d
b
y
ce
ll
u
lar
n
et
w
o
r
k
r
ec
o
n
f
i
g
u
r
atio
n
i
n
o
r
d
er
to
i
m
p
r
o
v
e
j
u
d
g
m
e
n
t
class
i
f
icatio
n
q
u
alit
y
.
I
n
t
h
is
s
ec
tio
n
,
p
r
e
-
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
es a
r
e
p
r
esen
ted
.
Ou
r
p
r
o
p
o
s
ed
m
e
th
o
d
w
h
ic
h
i
s
ap
p
licab
le
f
o
r
im
a
g
e
clas
s
i
f
ica
tio
n
an
d
th
e
al
g
o
r
it
h
m
is
d
is
c
u
s
s
ed
.
2
.1
.
P
re
pro
ce
s
s
ing
a
pp
ro
a
c
hes
T
h
e
o
b
j
ec
tiv
e
is
to
en
s
u
r
e
class
i
f
icatio
n
ac
cu
r
ac
y
(
C
A
)
,
p
r
ec
is
io
n
an
d
to
ac
ce
ler
ate
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
ti
m
e.
K
-
m
ea
n
s
,
Na
ïv
e
B
a
y
es
an
d
A
d
a
B
o
o
s
t
an
d
th
e
p
r
o
p
o
s
ed
class
if
ica
tio
n
m
o
d
els
ar
e
p
r
esen
ted
in
t
h
is
s
ec
tio
n
.
T
o
ex
ce
l
th
e
p
o
s
t
-
p
r
o
ce
s
s
i
n
g
co
m
p
u
tatio
n
,
t
h
e
r
ec
o
g
n
itio
n
m
o
d
el
a
s
p
r
ese
n
ted
i
n
[
8
]
is
u
s
ed
.
T
h
e
MO
A
s
i
m
u
latio
n
[
9
]
r
esu
lts
an
d
th
e
ir
p
er
f
o
r
m
an
ce
ar
e
e
v
alu
a
ted
.
2
.1
.1
.
K
-
M
ea
ns
cla
s
s
if
ica
t
io
n (
K
M
)
K
-
m
ea
n
s
cla
s
s
i
f
icatio
n
[
1
0
]
is
a
k
i
n
d
o
f
u
n
s
u
p
er
v
i
s
ed
class
if
ier
,
w
h
ic
h
is
e
m
p
lo
y
ed
as
d
ata
is
n
o
t
y
e
t
lab
el
ed
(
i.e
.
,
d
ata
w
it
h
o
u
t
g
r
o
u
p
s
)
.
T
h
e
alg
o
r
it
h
m
’
s
o
b
j
ec
tiv
e
i
s
to
clas
s
i
f
y
d
ef
i
n
ed
K
g
r
o
u
p
s
f
o
r
d
ata.
T
h
e
alg
o
r
ith
m
ca
lc
u
late
s
r
ep
ea
ted
ly
to
allo
ca
te
ea
ch
d
ata
t
o
o
n
e
o
f
v
ar
iab
le
K
g
r
o
u
p
s
b
ased
o
n
d
ata
ch
ar
ac
ter
is
tic
s
.
Data
is
cla
s
s
i
f
i
ed
b
ased
u
p
o
n
th
e
s
i
m
ilar
it
y
o
f
th
e
ir
ch
ar
ac
ter
is
tic
s
.
C
lo
s
e
s
t
w
it
h
K
g
r
o
u
p
s
(
K
-
m
ea
n
s
)
e
m
p
lo
y
ed
i
n
class
i
f
ica
tio
n
h
as
d
i
f
f
er
en
t
b
u
t
u
n
iq
u
e
f
u
n
ct
io
n
s
w
h
ic
h
d
if
f
er
f
r
o
m
o
t
h
er
alg
o
r
ith
m
s
.
I
t
i
s
u
n
s
u
p
er
v
is
ed
w
h
ic
h
r
eq
u
ir
es
n
o
i
n
p
u
t
p
r
o
b
ab
ilit
y
d
en
s
it
y
f
u
n
ctio
n
.
T
h
i
s
K
-
m
ea
n
s
is
a
lag
g
i
n
g
lear
n
in
g
alg
o
r
ith
m
,
w
h
ic
h
co
m
p
u
tes
d
ata
d
u
r
in
g
t
h
e
test
in
g
p
er
io
d
,
r
ath
er
th
a
n
i
n
t
h
e
lear
n
i
n
g
p
h
ase.
A
b
en
e
f
it
o
f
K
-
m
ea
n
s
is
t
h
at
it
q
u
ic
k
l
y
ad
ap
t
s
an
y
alter
atio
n
s
.
B
u
t
a
d
r
a
w
b
ac
k
is
th
e
co
m
p
u
tatio
n
a
l
co
s
t
d
u
e
to
s
tate
s
p
ac
e
co
m
p
le
x
it
y
.
2
.1
.2
.
Na
ïv
e
B
a
y
es c
la
s
s
if
ica
t
io
n (
NB
)
T
h
e
Naiv
e
B
a
y
es
C
la
s
s
i
f
icati
o
n
[
1
1
]
b
ased
o
n
th
e
B
a
y
esia
n
t
h
eo
r
y
is
ap
p
r
o
p
r
iate
f
o
r
au
to
n
o
m
o
u
s
in
p
u
t
v
ar
iab
les.
R
eg
ar
d
les
s
o
f
its
in
co
m
p
lex
i
t
y
a
n
d
lo
w
co
m
p
u
tatio
n
al
co
s
t,
NB
ca
n
o
u
tp
er
f
o
r
m
m
o
r
e
ad
v
an
ce
d
cla
s
s
i
f
icatio
n
.
NB
class
i
f
ier
s
ca
n
lev
er
a
n
u
m
b
e
r
o
f
i
n
d
ep
en
d
en
t
v
ar
iab
les
wh
eth
er
clas
s
i
f
ied
o
r
r
ep
ea
ted
.
Giv
en
a
s
e
t
o
f
d
i
m
e
n
s
io
n
al
attr
ib
u
te
v
ec
to
r
o
f
X
=
{x
1
,x
2
,x
3
,...,x
d
},
a
n
d
th
e
s
u
b
s
eq
u
en
t
p
r
o
b
ab
ilit
y
f
o
r
th
e
e
v
en
t
C
j
a
m
o
n
g
al
l
p
o
s
s
ib
le
o
u
tco
m
e
s
C
=
{c
1
,c
2
,c
3
,...,c
d
},
in
a
co
n
v
en
t
io
n
al
la
n
g
u
a
g
e,
X
is
ca
lled
th
e
p
r
ed
icto
r
s
an
d
C
is
t
h
e
s
et
o
f
d
if
f
er
e
n
t
clas
s
es
p
r
ese
n
ted
in
t
h
e
d
ep
en
d
en
t
v
ar
iab
le.
Ass
u
m
e
x
d
ca
n
tak
e
d
if
f
er
e
n
t
C
j
v
al
u
es,
n
a
m
el
y
,
P
(
C
j
/X)
>
P
(
C
k
/X)
f
o
r
1
≤
k
≤
d
an
d
k
≠
j.
T
h
e
NB
cla
s
s
i
fi
er
co
m
p
u
te
s
a
p
r
o
b
a
b
ilit
y
o
f
C
j
a
s
f
o
llo
w
i
n
g
P
(
C
j
/X)
=
P(X
/C
j
)
P
(
C
j
)
/
P(
X)
.
T
h
e
v
alu
e
s
P
(
X/C
j
)
an
d
P
(
X)
ar
e
esti
m
ated
f
r
o
m
t
h
e
tr
ain
i
n
g
.
T
h
e
NB
alg
o
r
ith
m
i
s
s
h
o
w
n
i
n
Fi
g
u
r
e
1.
A
l
g
o
r
i
t
h
m:
NB
R
e
q
u
i
r
e
:
D
a
t
a
ma
t
r
i
x
[
D
]
xy
w
i
t
h
x
r
o
w
s a
n
d
y
c
o
l
u
m
n
s
f
o
r
p
=
1
t
o
x
do
f
o
r
k
=
1
t
o
y
do
C
o
n
st
r
u
c
t
a
f
r
e
q
u
e
n
c
y
t
a
b
l
e
f
o
r
a
l
l
c
h
a
r
a
c
t
e
r
i
st
i
c
s
f
o
r
C
p
B
u
i
l
d
t
h
e
p
r
o
sp
e
c
t
t
a
b
l
e
f
o
r
a
l
l
c
h
a
r
a
c
t
e
r
i
st
i
c
s fo
r
C
p
C
a
l
c
u
l
a
t
e
t
h
e
c
o
n
d
i
t
i
o
n
a
l
p
r
o
b
a
b
i
l
i
t
y
f
o
r
C
p
C
a
l
c
u
l
a
t
e
t
h
e
max
i
m
u
m
p
r
o
b
a
b
i
l
i
t
y
f
o
r
C
p
e
n
d
f
o
r
e
n
d
f
o
r
Fig
u
r
e
1
.
NB
alg
o
r
ith
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
r
o
p
o
s
ed
a
lg
o
r
ith
m
fo
r
ima
g
e
cla
s
s
i
fica
tio
n
u
s
in
g
r
eg
r
ess
io
n
-
b
a
s
ed
…
(
C
h
a
n
in
to
r
n
Jitta
w
ir
iya
n
u
ko
o
n
)
1023
2
.1
.3
.
Ada
B
o
o
s
t
cla
s
s
if
ica
t
i
o
n (
AB
)
A
d
a
B
o
o
s
t
(
A
B
)
al
g
o
r
ith
m
[
1
2
]
r
ep
air
s
d
elica
te
to
a
to
u
g
h
l
ea
r
n
in
g
e
n
v
ir
o
n
m
e
n
t.
T
h
e
w
ei
g
h
t
d
iv
id
e
s
th
e
d
ata
m
atr
i
x
D
xy
in
to
2
p
ar
ts
s
y
m
m
etr
icall
y
.
First
to
u
g
h
p
ar
t
o
f
th
e
w
eig
h
t
i
s
s
et
to
b
e
th
e
p
er
f
ec
t
class
if
ied
p
ar
t,
an
d
th
e
d
elica
te
p
ar
t
is
allo
ca
t
ed
t
o
th
e
n
o
n
-
cla
s
s
i
f
i
ed
p
ar
t.
T
h
e
P
o
is
s
o
n
d
is
tr
ib
u
tio
n
f
u
n
ctio
n
f
o
r
ca
lcu
lati
n
g
th
e
r
a
n
d
o
m
p
r
o
b
ab
ilit
y
i
n
o
r
d
er
to
class
if
y
t
h
e
d
ata
m
o
d
el
h
as
o
p
ted
.
T
h
e
id
ea
o
f
A
B
is
to
ag
r
ee
o
n
a
s
er
ies
o
f
d
elica
te
lear
n
er
s
.
T
h
e
w
ei
g
h
ted
v
ar
iab
le
i
s
d
e
s
ig
n
ed
to
a
d
ata
m
o
d
e
l
w
h
ich
is
m
i
s
clas
s
i
f
ied
i
n
th
e
p
r
ev
io
u
s
r
ep
etitio
n
.
O
n
l
y
th
e
p
r
ese
n
t
th
e
w
ei
g
h
t
in
g
v
ar
iab
le
ch
an
g
es
ac
co
r
d
in
g
to
th
e
A
B
w
ei
g
h
t
as
p
r
o
ce
ed
in
g
th
r
o
u
g
h
ea
c
h
iter
atio
n
o
f
ca
lcu
latio
n
.
T
h
e
ap
p
r
o
x
i
m
atio
n
m
o
v
e
s
o
n
w
it
h
it
er
ativ
el
y
co
m
p
u
ti
n
g
th
r
o
u
g
h
t
h
e
w
ei
g
h
ted
class
i
f
ic
atio
n
u
n
til t
h
e
ter
m
in
al
r
o
u
n
d
.
T
h
e
alg
o
r
ith
m
i
s
d
ep
icted
in
F
ig
u
r
e
2.
A
l
g
o
r
i
t
h
m:
AB
R
e
q
u
i
r
e
:
D
a
t
a
m
a
t
r
i
x
[
D
]
xy
w
i
t
h
x
r
o
w
s a
n
d
y
c
o
l
u
m
n
s
E
n
su
r
e
:
[
D]
xy
→
[
D
1
]
a
n
d
[
D
2
]
,
Q
=
d
i
me
n
si
o
n
o
f
[
D
]
S
e
t
:
W
e
i
g
h
t
v
a
r
i
a
b
l
e
i
s
i
n
i
t
i
a
l
l
y
se
t
t
o
b
e
w
q
(
=
1
/
Q
)
f
o
r
i
=
1
t
o
x
do
f
o
r
j
=
1
t
o
y
do
f
o
r
k
=
1
t
o
K
do
T
a
k
e
C
k
(
a
)
a
f
t
e
r
mi
n
i
m
i
z
i
n
g
e
r
r
o
r
o
f
w
e
i
g
h
t
v
a
r
i
a
b
l
e
E
k
C
a
l
c
u
l
a
t
e
E
k
=
∑
(
)
(
)
C
a
l
c
u
l
a
t
e
α
k
=
∑
(
)
(
)
∑
(
)
C
a
l
c
u
l
a
t
e
β
k
=
l
o
g
e
(
)
U
si
n
g
P
o
i
sso
n
d
i
s
t
r
i
b
u
t
i
o
n
f
u
n
c
t
i
o
n
t
o
r
a
n
d
o
mi
z
e
a
n
d
u
p
d
a
t
e
t
h
e
w
e
i
g
h
t
v
a
r
i
a
b
l
e
(
)
=
(
)
*
(
)
+
e
n
d
f
o
r
A
p
p
r
o
x
i
mat
e
t
h
r
o
u
g
h
f
i
n
a
l
r
o
u
n
d
Y
K
(
a
)
=
sg
n
∑
(
)
{
-
1
,
0
,
1
}
e
n
d
f
o
r
e
n
d
f
o
r
Fig
u
r
e
2
.
A
B
alg
o
r
it
h
m
2
.
2
.
P
r
o
po
s
ed
a
lg
o
rit
h
m
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
i
s
b
ased
u
p
o
n
a
lo
g
is
tic
r
eg
r
es
s
i
o
n
-
b
ased
lear
n
i
n
g
en
v
ir
o
n
m
e
n
t
w
h
ic
h
in
te
g
r
ates
m
u
ltip
le
cla
s
s
i
f
icat
i
o
n
s
i
n
o
r
d
er
to
m
a
x
i
m
ize
t
h
e
o
b
s
er
v
ed
p
r
o
b
ab
ilit
y
f
i
g
u
r
e.
A
t
th
e
lo
w
le
v
el
o
f
co
m
p
u
tatio
n
,
th
er
e
ar
e
m
is
ce
ll
an
eo
u
s
lear
n
i
n
g
al
g
o
r
ith
m
s
wh
ich
ar
e
tr
ain
ed
i
n
d
ep
en
d
en
tl
y
.
T
h
is
is
n
o
t
s
i
m
ilar
to
o
th
er
alg
o
r
ith
m
s
w
h
ic
h
ta
k
e
th
e
s
a
m
p
le
v
al
u
es
th
a
t
m
in
i
m
ize
th
e
to
tal
o
f
m
ea
n
s
q
u
ar
ed
er
r
o
r
s
.
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
d
ea
ls
w
ith
th
e
g
r
o
u
p
in
g
o
f
p
r
e
-
p
r
o
ce
s
s
in
g
tec
h
n
iq
u
e
s
f
o
r
t
h
e
p
o
s
t
-
p
r
o
ce
s
s
i
n
g
o
f
t
h
e
o
u
tco
m
e
at
h
i
g
h
er
lear
n
in
g
l
ev
el.
No
tice
t
h
at
t
h
e
o
r
ig
i
n
al
lear
n
in
g
e
n
v
ir
o
n
m
e
n
t
i
s
n
o
t
tailo
r
ed
w
h
ile
t
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
tar
g
et
s
at
ac
h
iev
ab
le
h
i
g
h
er
ac
cu
r
ac
y
(
a
s
w
ell
as
p
r
ec
is
io
n
)
i
n
t
h
e
cla
s
s
i
f
icatio
n
o
f
d
ata.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
is
i
ter
a
tiv
el
y
tr
ain
ed
t
h
r
o
u
g
h
th
e
o
u
t
co
m
e
s
f
r
o
m
t
h
e
lo
w
le
v
el
o
f
co
m
p
u
tatio
n
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
i
s
lis
ted
in
Fig
u
r
e
3.
P
r
o
p
o
s
e
d
A
l
g
o
r
i
t
h
m
R
e
q
u
i
r
e
:
D
a
t
a
ma
t
r
i
x
[
D
]
xy
w
i
t
h
x
r
o
w
s a
n
d
y
c
o
l
u
m
n
s
E
n
su
r
e
:
[
D
]
x
y
,
Z
=
n
u
m
b
e
r
o
f
c
l
a
ssi
f
i
e
r
s,
Q
=
d
i
me
n
si
o
n
o
f
[
D
]
f
o
r
I
=
1
t
o
x
do
f
o
r
j
=
1
t
o
y
do
f
o
r
k
=
1
to
P
do
/
*
*
L
o
w
l
e
v
e
l
c
o
mp
u
t
a
t
i
o
n
*
*
/
L
e
a
r
n
e
r
Z
k
a
p
p
l
y
i
n
g
f
o
r
d
a
t
a
ma
t
r
i
x
D
e
n
d
f
o
r
f
o
r
n
=
1
to
Q
do
/
*
*
R
e
g
r
e
ssi
o
n
-
b
a
se
d
c
o
m
p
u
t
a
t
i
o
n
t
o
max
i
m
i
z
e
p
r
o
b
a
b
i
l
i
t
y
*
*
/
D
z
=
{
x’
n
,y
n
},
i
n
w
h
i
c
h
x’
n
=
z
0
+
z
1
x
n
+
z
2
x
n
+
.
.
.
+
z
P
x
n
e
n
d
f
o
r
A
p
p
l
y
l
e
a
r
n
e
r
Z
w
i
t
h
D
z
/
*
*
H
i
g
h
l
e
v
e
l
c
o
mp
u
t
a
t
i
o
n
*
*
/
R
e
t
u
r
n
Z
e
n
d
f
o
r
e
n
d
f
o
r
Fig
u
r
e
3
.
P
r
o
p
o
s
ed
alg
o
r
ith
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
0
2
1
-
1027
1024
3.
RE
CO
G
NI
T
I
O
N
M
O
DE
L
S
T
o
s
ea
r
ch
d
atab
ases
o
f
d
ig
ital
i
m
a
g
es
i
s
a
p
r
o
v
o
k
i
n
g
j
o
b
p
ar
t
icu
lar
l
y
f
o
r
i
m
a
g
e
d
ata
r
etr
iev
al
.
Sear
c
h
en
g
i
n
e
s
co
m
p
u
te
th
e
m
atc
h
b
et
w
ee
n
th
e
q
u
er
y
i
m
a
g
e
an
d
all
i
m
ag
e
s
s
to
r
ed
in
th
e
d
atab
ase
an
d
class
i
f
y
t
h
e
s
eq
u
en
ce
o
f
i
m
ag
e
s
d
u
e
to
t
h
eir
m
atc
h
es.
O
n
e
d
r
a
w
b
ac
k
is
t
h
at
ti
m
e
to
s
o
r
t
all
m
atc
h
ed
i
m
a
g
es
r
eq
u
ir
es
O(
lo
g
d
!
)
in
ca
s
e
o
f
d
d
ata
i
m
ag
e
s
.
T
o
tr
ain
to
r
ec
o
g
n
ize
i
m
a
g
es
r
eq
u
ir
es
n
u
m
er
o
u
s
i
m
ag
es.
T
h
is
in
c
lu
d
e
s
in
p
u
tt
in
g
s
ev
er
al
m
illi
o
n
i
m
ag
e
s
i
n
to
d
ee
p
lear
n
in
g
lev
el
[
1
3
]
.
I
n
b
et
w
ee
n
,
t
h
e
p
r
o
ce
s
s
p
r
o
d
u
ce
s
ch
ar
ac
ter
is
tic
s
f
r
o
m
t
h
e
d
ig
ital
i
m
ag
e.
T
h
ese
ch
ar
ac
ter
is
tics
ar
e
j
u
s
t
th
e
o
u
tlin
e
s
o
r
i
m
ag
e
ex
tr
ac
tio
n
.
T
h
e
id
ea
is
th
a
t
th
o
s
e
ch
ar
ac
ter
is
tics
ca
n
d
ig
d
ee
p
er
d
o
w
n
to
an
o
t
h
er
la
y
er
,
f
o
r
in
s
ta
n
ce
,
th
e
d
i
s
tan
c
e
v
ec
to
r
f
r
o
m
ea
c
h
class
o
r
th
e
d
en
s
i
t
y
o
f
ea
c
h
ch
ar
ac
ter
is
tic.
On
ce
t
h
ese
ch
ar
ac
ter
is
tics
ar
e
s
at
u
r
ated
th
e
n
o
th
er
r
ec
o
g
n
iti
o
n
j
o
b
s
o
r
th
e
i
m
a
g
e
m
atch
in
g
r
eq
u
ir
e
s
o
n
l
y
m
ac
h
i
n
e
lear
n
in
g
le
v
el.
I
t
is
p
o
s
i
tiv
e
to
m
ak
e
u
s
e
o
f
a
n
al
g
o
r
ith
m
w
h
ic
h
h
as
alr
ea
d
y
b
ee
n
tr
ai
n
ed
o
n
lo
ts
o
f
i
m
a
g
es
w
h
ile
a
co
llectio
n
an
d
a
tr
ain
i
n
g
o
f
s
e
v
er
al
m
il
lio
n
i
m
ag
e
s
i
s
n
e
x
t
to
i
m
p
o
s
s
ib
le.
Fo
r
in
s
tan
ce
,
R
esNet
[
1
4
]
is
a
d
ee
p
lear
n
in
g
m
o
d
el
tr
ai
n
ed
f
o
r
i
m
ag
e
r
ec
o
g
n
i
tio
n
i
n
co
m
p
u
ter
v
is
io
n
.
W
it
h
b
u
il
t
-
in
f
u
n
cti
o
n
,
th
e
s
o
f
t
w
ar
e
ca
n
g
e
n
er
ate
a
m
o
d
el
w
i
th
a
v
ec
to
r
r
ep
r
esen
tatio
n
o
f
4
,
0
9
6
ch
ar
ac
ter
is
tics
b
et
w
ee
n
ze
r
o
an
d
o
n
e
f
r
o
m
in
p
u
t
i
m
ag
es.
T
h
e
i
m
a
g
e
r
ec
o
g
n
i
tio
n
m
o
d
el
b
ased
u
p
o
n
th
ese
ch
ar
ac
ter
i
s
tic
v
ec
to
r
s
is
d
ev
elo
p
ed
w
i
th
o
u
t
th
e
w
o
r
k
lo
ad
o
f
tr
ain
in
g
at
d
ee
p
lear
n
in
g
lev
el
o
n
to
m
as
s
iv
e
d
ata
i
m
a
g
es.
I
n
t
h
i
s
r
esear
c
h
,
t
h
e
o
p
en
-
s
o
u
r
ce
b
ased
s
i
m
u
l
atio
n
to
o
l,
MO
A
is
u
s
ed
f
o
r
t
h
e
i
m
a
g
e
a
n
al
y
tics
.
T
w
e
n
t
y
d
i
g
ital
i
m
a
g
es
i
n
th
e
d
atab
ase
h
av
e
b
ee
n
ch
o
s
e
n
.
T
h
e
ex
p
er
im
e
n
t
h
as
b
ee
n
e
x
ec
u
ted
o
n
a
n
A
s
u
s
W
in
d
o
w
s
7
w
it
h
I
n
tel® Co
r
e
™
i
7
C
P
U,
2
.
2
GHz
P
r
o
ce
s
s
o
r
an
d
8
GB
R
A
M
o
n
b
o
ar
d
.
T
h
e
i
m
a
g
es h
a
v
e
b
ee
n
ch
o
s
en
in
o
r
d
er
th
at
t
h
e
y
a
r
e
all
d
if
f
er
e
n
t
i
n
s
ize,
n
u
m
b
er
o
f
attr
ib
u
t
es,
i
n
s
ta
n
ce
s
,
an
d
co
n
te
n
ts
.
T
h
e
ex
p
er
i
m
e
n
tal
m
o
d
el
i
s
g
iv
e
n
i
n
Fi
g
u
r
e
4
.
I
n
o
r
d
er
to
ev
alu
at
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
f
o
r
an
i
m
p
r
o
v
e
m
e
n
t
o
f
CA
a
n
d
p
r
ec
is
io
n
is
also
d
en
o
ted
by
t
h
e
p
r
e
-
p
r
o
ce
s
s
in
g
r
es
u
lt
s
f
r
o
m
r
an
d
o
m
l
y
s
elec
ted
f
o
u
r
i
m
ag
e
s
as p
r
esen
ted
i
n
T
ab
le
1
.
Fig
u
r
e
4
.
E
x
p
er
i
m
en
tal
m
o
d
el
T
ab
le
1
.
R
esu
lts
f
r
o
m
P
r
e
-
p
r
o
ce
s
s
i
n
g
E
v
al
u
atio
n
I
mag
e
2
C
l
a
ssi
f
i
e
r
T
y
p
e
C
A
(
%)
P
r
e
c
i
si
o
n
(
%)
KM
5
2
.
2
4
7
.
2
NB
5
0
.
2
7
4
.
3
AB
7
3
.
9
7
5
.
2
P
r
o
p
o
se
d
8
2
.
6
8
4
.
3
I
mag
e
6
C
l
a
ssi
f
i
e
r
T
y
p
e
C
A
(
%)
P
r
e
c
i
si
o
n
(
%)
KM
4
1
.
9
3
7
.
2
NB
2
2
.
6
3
8
.
3
AB
5
1
.
6
4
2
.
2
P
r
o
p
o
se
d
5
8
.
4
6
2
.
4
I
mag
e
1
1
C
l
a
ssi
f
i
e
r
T
y
p
e
C
A
(
%)
P
r
e
c
i
si
o
n
(
%)
KM
5
3
.
6
43
NB
1
7
.
9
3
.
2
AB
4
6
.
4
3
5
.
9
P
r
o
p
o
se
d
6
0
.
7
4
6
.
9
I
mag
e
1
4
C
l
a
ssi
f
i
e
r
T
y
p
e
C
A
(
%)
P
r
e
c
i
si
o
n
(
%)
KM
48
5
0
.
5
NB
36
3
6
.
9
AB
44
67
P
r
o
p
o
se
d
68
8
4
.
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
r
o
p
o
s
ed
a
lg
o
r
ith
m
fo
r
ima
g
e
cla
s
s
i
fica
tio
n
u
s
in
g
r
eg
r
ess
io
n
-
b
a
s
ed
…
(
C
h
a
n
in
to
r
n
Jitta
w
ir
iya
n
u
ko
o
n
)
1025
I
n
th
e
r
ec
o
g
n
i
tio
n
m
o
d
el,
th
e
s
p
ee
d
o
f
m
atc
h
i
n
g
ca
lc
u
lat
io
n
d
ep
en
d
s
o
n
its
P
(
n
)
w
h
ic
h
i
s
t
h
e
n
th
p
ix
e
l
m
o
v
e
s
w
it
h
v
elo
cit
y
(
v
)
.
I
n
Fig
u
r
e
5
,
o
n
e
p
ix
el
m
o
v
es
at
a
v
elo
cit
y
o
f
v
at
t
.
T
h
e
p
o
s
itio
n
o
f
th
e
m
atc
h
in
g
p
ix
el
in
t
h
e
s
ta
t
e
s
p
ac
e
at
t +
1
ca
n
b
e
g
iv
e
n
b
y
E
q
u
at
io
n
(
1
)
.
Fig
u
r
e
5
.
Ma
tch
in
g
c
h
ec
k
o
f
t
h
e
p
ix
el
s
in
t
h
e
s
tate
s
p
ac
e
P
n+
1
(
t+1
)
=
P
n
(
t
)
+
v
(
1
)
L
et
ῡ
b
e
t
h
e
a
v
er
ag
e
v
elo
cit
y
an
d
q
b
e
th
e
d
i
m
e
n
s
io
n
o
f
a
d
ig
ital
i
m
ag
e
th
e
n
t
h
e
co
m
p
u
ta
tio
n
al
co
s
t
f
o
r
r
ec
o
g
n
iz
in
g
a
n
i
m
a
g
e
is
O(
q
)
.
T
w
en
t
y
d
ig
ita
l
i
m
a
g
es
in
t
h
e
d
atab
ase
as
lis
ted
in
T
ab
le
2
h
av
e
b
ee
n
e
m
p
lo
y
ed
b
y
p
o
s
t
-
p
r
o
ce
s
s
with
d
i
f
f
er
en
t
e
m
b
ed
d
in
g
al
g
o
r
ith
m
s
(
G
L
N,
I
L
S
VR
C
a
n
d
C
P
VR
)
i
n
o
r
d
er
to
r
ec
o
g
n
iz
e
t
h
e
m
atc
h
es.
A
p
r
o
ce
s
s
i
n
g
f
o
r
r
ed
u
cin
g
o
f
s
ta
te
s
p
a
ce
p
r
o
b
lem
an
d
ti
m
e
is
d
e
m
o
n
s
tr
ated
in
[
1
5
]
.
Ta
b
le
2
.
Featu
r
e
o
f
t
w
e
n
t
y
i
m
ag
es
I
mag
e
D
i
me
n
si
o
n
(
p
i
x
e
l
s)
S
i
z
e
(
K
B
)
1
2
5
5
x
1
9
8
9
.
0
2
5
2
2
7
5
x
1
8
3
8
.
1
4
2
3
2
0
0
x
2
3
1
7
.
2
2
3
4
2
7
5
x
1
8
3
5
.
9
0
6
5
4
0
0
x
3
2
5
8
0
.
4
2
0
6
4
0
0
x
2
9
4
7
6
.
8
7
0
7
2
5
9
x
1
9
4
9
.
9
8
4
8
2
9
0
x
1
7
8
1
2
.
9
5
9
9
4
0
0
x
2
2
6
8
8
.
8
9
4
10
4
0
0
x
3
2
6
1
1
7
.
3
6
6
11
4
0
0
x
3
1
7
1
0
3
.
1
7
7
12
1
0
2
4
x
8
4
9
7
2
6
.
2
5
1
13
3
3
5
x
1
5
1
8
.
9
0
8
14
4
0
0
x
2
8
6
1
0
2
.
3
8
2
15
3
1
8
x
1
5
9
1
0
.
9
9
5
16
4
0
0
x
2
6
1
9
6
.
2
2
6
17
3
8
4
x
2
4
7
1
7
.
3
6
0
18
4
0
0
x
2
6
9
4
1
.
7
1
7
19
2
2
5
x
2
2
5
6
.
0
8
9
20
2
2
9
x
2
2
0
1
0
.
1
3
5
4.
R
ES
U
L
T
S AN
D
AN
AL
Y
SI
S
T
h
e
co
m
p
u
tatio
n
o
f
th
e
i
m
ag
e
m
atc
h
i
n
g
b
ased
o
n
th
e
in
p
u
t
p
ar
a
m
eter
s
f
r
o
m
p
r
e
-
p
r
o
ce
s
s
in
g
alg
o
r
ith
m
s
a
n
d
th
eir
f
i
n
al
r
es
u
lts
i
n
cl
u
d
in
g
m
atc
h
p
er
ce
n
ta
g
e
an
d
ac
ce
p
tan
ce
r
an
g
e
o
f
h
ig
h
(
h
i:
ab
o
v
e
5
1
)
,
m
ed
iu
m
(
m
ed
:
b
et
w
ee
n
2
5
an
d
5
0
)
an
d
lo
w
(
lo
:
lo
w
er
th
an
2
4
)
.
I
m
a
g
e
n
u
m
b
er
o
n
e
i
s
u
s
ed
as
a
r
ef
er
e
n
ce
i
m
a
g
e
i
n
t
h
e
r
ec
o
g
n
i
tio
n
m
o
d
el
an
d
in
f
in
d
i
n
g
f
o
r
an
o
t
h
er
t
w
o
m
atc
h
es.
T
h
e
r
etr
iev
al
s
i
m
ilar
itie
s
(
m
atc
h
in
g
p
er
ce
n
tag
e)
f
o
r
th
e
GL
N
e
m
b
ed
d
in
g
al
g
o
r
ith
m
s
w
it
h
t
h
e
f
o
u
r
p
r
e
-
p
r
o
ce
s
s
i
n
g
m
et
h
o
d
s
ar
e
s
u
m
m
ar
is
ed
i
n
T
ab
le
3
.
A
s
s
ee
n
th
e
m
atc
h
i
n
g
p
er
ce
n
tag
e
f
o
r
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
is
h
i
g
h
er
th
a
n
o
th
er
s
,
b
u
t a
f
e
w
p
er
ce
n
ta
g
e
p
o
in
ts
.
T
h
e
d
if
f
er
en
ce
in
t
h
is
p
er
ce
n
tag
e
h
elp
s
r
ed
u
ce
a
clar
if
icatio
n
in
d
ec
is
io
n
m
ak
i
n
g
f
o
r
im
a
g
e
s
i
m
ilar
it
y
.
R
es
u
lts
f
r
o
m
t
h
e
I
L
S
VR
C
-
2
0
1
4
alg
o
r
ith
m
ar
e
also
s
ee
n
i
n
t
h
e
id
en
tical
d
ir
ec
tio
n
.
T
h
e
r
esu
lts
o
b
t
ain
ed
b
y
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
ar
e
h
ig
h
er
in
m
a
tch
i
n
g
p
er
ce
n
ta
g
e.
A
m
o
r
e
th
o
r
o
u
g
h
e
v
alu
a
tio
n
u
s
i
n
g
C
P
VR
-
2015
alg
o
r
ith
m
co
n
f
ir
m
s
th
e
r
es
u
l
ts
o
b
tain
ed
f
r
o
m
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
s
till
o
u
tp
er
f
o
r
m
.
T
h
e
p
er
f
o
r
m
an
ce
s
d
ep
icted
in
T
ab
le
3
ar
e
s
i
m
ilar
to
th
e
co
r
r
esp
o
n
d
in
g
r
es
u
lts
i
n
T
ab
le
1
d
em
o
n
s
tr
ati
n
g
t
h
at
t
h
er
e
i
s
n
o
d
eg
r
ad
atio
n
i
n
t
h
e
r
etr
ie
v
al
p
e
r
f
o
r
m
an
ce
w
h
en
th
e
p
r
e
-
p
r
o
ce
s
s
i
s
c
h
o
s
e
n
i
n
o
r
d
er
to
clas
s
if
y
i
m
a
g
es
f
r
o
m
t
h
e
d
atab
ase.
No
t
to
m
en
tio
n
f
o
r
all
t
h
r
ee
d
if
f
er
en
t
i
n
v
esti
g
atio
n
s
,
th
e
p
r
o
p
o
s
ed
m
e
th
o
d
r
es
u
lts
i
n
a
m
id
-
r
a
n
g
e
-
ac
ce
p
tan
ce
lev
el
w
h
ile
o
t
h
er
s
g
iv
e
o
n
l
y
lo
w
-
r
an
g
e
-
le
v
el.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2088
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
2
,
A
p
r
il 2
0
1
9
:
1
0
2
1
-
1027
1026
T
ab
le
3
.
R
e
s
u
lts
o
f
t
w
en
t
y
i
m
a
g
es
m
atc
h
in
g
P
r
e
p
r
o
c
e
ss
Emb
e
d
d
i
n
g
I
mag
e
N
o
.
M
a
t
c
h
i
n
g
(
%)
A
c
c
e
p
t
a
n
c
e
KM
G
L
N
19
2
1
3
.
4
9
1
8
.
2
2
Lo
Lo
NB
10
19
1
3
.
8
1
7
.
0
6
Lo
Lo
AB
15
19
2
3
.
3
2
2
6
.
8
5
Lo
M
e
d
P
r
o
p
o
se
d
20
2
2
8
.
9
2
9
.
4
5
M
e
d
M
e
d
KM
I
L
S
V
R
C
-
2
0
1
4
15
16
1
4
.
5
2
3
.
7
Lo
Lo
NB
15
16
1
5
.
5
1
1
8
.
0
7
Lo
Lo
AB
11
16
1
8
.
0
7
2
3
.
5
Lo
Lo
P
r
o
p
o
se
d
5
9
2
7
.
0
2
7
.
6
2
M
e
d
M
e
d
KM
C
P
V
R
-
2
0
1
5
20
15
1
4
.
9
4
1
6
.
6
Lo
Lo
NB
2
9
1
5
.
5
1
1
6
.
4
7
Lo
Lo
AB
2
4
1
8
.
3
1
1
9
.
9
6
Lo
Lo
P
r
o
p
o
se
d
5
9
2
7
.
0
4
3
0
.
1
M
e
d
M
e
d
5.
CO
NCLU
SI
O
NS A
ND
F
UT
URE WO
RK
T
h
e
r
esu
lts
o
b
tai
n
ed
w
it
h
t
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
d
e
m
o
n
s
tr
ate
t
h
at
p
r
e
-
p
r
o
ce
s
s
i
n
c
r
ea
s
es
t
h
e
class
i
f
icatio
n
ac
cu
r
ac
y
an
d
p
r
ec
is
io
n
w
it
h
o
u
t
s
ac
r
i
f
ici
n
g
t
h
e
am
o
u
n
t
o
f
r
eq
u
ir
ed
m
atc
h
i
n
g
co
m
p
u
tatio
n
.
T
h
e
p
r
o
p
o
s
ed
tech
n
iq
u
e
ca
n
b
e
u
s
ed
f
o
r
a
s
ca
lab
le
d
i
g
ital
i
m
a
g
e
f
r
o
m
lar
g
e
d
atab
ase
s
.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
o
u
tp
er
f
o
r
m
s
o
th
er
t
h
r
ee
al
g
o
r
i
th
m
s
in
th
e
p
r
e
-
p
r
o
ce
s
s
i
n
g
p
h
ase,
ev
e
n
t
h
o
u
g
h
o
n
l
y
m
ar
g
i
n
all
y
i
n
a
ll
ca
s
es.
I
n
th
e
p
o
s
t
-
p
r
o
ce
s
s
in
g
p
h
a
s
e,
th
r
ee
alg
o
r
ith
m
s
n
a
m
el
y
(
G
L
N,
I
L
SV
R
C
a
n
d
C
P
VR
)
ar
e
ap
p
l
ied
f
o
r
r
ec
o
g
n
iz
i
n
g
i
m
a
g
e
s
i
m
ilar
it
y
.
I
n
th
e
p
o
s
t
-
p
r
o
ce
s
s
in
g
p
h
ase,
r
es
u
lt
s
f
r
o
m
p
r
o
p
o
s
ed
m
et
h
o
d
also
m
ar
g
in
a
ll
y
i
m
p
r
o
v
e
t
h
e
m
atc
h
in
g
p
er
ce
n
tag
e.
Mo
r
e
s
o
p
h
is
ticated
s
i
m
ilar
it
y
m
ea
s
u
r
e
s
[
1
6
]
w
h
ic
h
h
av
e
b
ee
n
u
s
ed
i
n
t
h
e
v
id
eo
s
tr
ea
m
ar
e
b
ein
g
c
u
r
r
en
tl
y
i
n
v
esti
g
at
ed
an
d
th
e
s
e
r
esu
lts
w
ill
b
e
p
r
esen
ted
i
n
t
h
e
n
ea
r
f
u
tu
r
e.
An
o
th
er
d
i
r
ec
tio
n
o
f
f
u
tu
r
e
w
o
r
k
is
to
id
en
tify
a
b
en
ch
m
ar
k
ag
ai
n
s
t
w
h
ic
h
t
h
e
d
if
f
er
en
t
m
atc
h
i
n
g
r
an
g
e
s
ca
n
b
e
s
et.
T
h
e
ex
ec
u
tio
n
ti
m
e
i
n
ea
ch
p
h
a
s
e
w
ill b
e
tak
en
in
to
ac
co
u
n
t a
s
w
e
ll.
RE
F
E
R
E
NC
E
S
[1
]
N.
Dh
a
n
a
c
h
a
n
d
ra
,
Y.
J
.
Ch
a
n
u
a
n
d
K.
M
a
n
g
lem
,
“
I
m
a
g
e
S
e
g
m
e
n
tatio
n
Us
in
g
K
-
m
e
a
n
s
Clu
ste
rin
g
A
l
g
o
rit
h
m
a
n
d
S
u
b
trac
ti
v
e
Clu
ste
ri
n
g
A
lg
o
rit
h
m
,
”
Pro
c
e
d
ia
C
o
mp
u
ter
S
c
ien
c
e
,
v
o
l.
5
4
,
p
p
.
7
6
4
-
7
7
1
,
2
0
1
5
.
[2
]
Y.
F
a
rh
a
n
g
,
“
F
a
c
e
E
x
trac
ti
o
n
f
r
o
m
I
m
a
g
e
b
a
se
d
o
n
K
-
M
e
a
n
s
C
lu
ste
rin
g
A
lg
o
rit
h
m
s,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
Ad
v
a
n
c
e
d
Co
mp
u
ter
S
c
ien
c
e
a
n
d
Ap
p
li
c
a
ti
o
n
s (
IJ
ACS
A),
v
o
l
.
8
,
n
o
.
9
,
p
p
.
9
6
-
1
0
7
,
2
0
1
7
.
[3
]
P
.
X
i
n
a
n
d
H.
S
a
g
a
n
,
“
Dig
it
a
l
I
m
a
g
e
Clu
ste
rin
g
A
l
g
o
rit
h
m
b
a
se
d
o
n
M
u
lt
i
-
a
g
e
n
t
Ce
n
ter
Op
ti
m
iza
ti
o
n
,
”
J
o
u
r
n
a
l
o
f
Dig
it
a
l
In
f
o
rm
a
ti
o
n
M
a
n
a
g
e
me
n
t
(
IJ
DI
M
),
v
o
l.
1
4
,
n
o
.
1
,
p
p
.
8
-
1
4
,
2
0
1
6
.
[4
]
X
.
Z
h
a
o
,
Y.
L
i
a
n
d
Q.
Zh
a
o
,
“
M
a
h
a
lan
o
b
is
d
istan
c
e
b
a
se
d
o
n
f
u
z
z
y
c
lu
ste
rin
g
a
l
g
o
rit
h
m
f
o
r
i
m
a
g
e
se
g
m
e
n
tatio
n
,
”
Dig
it
a
l
S
i
g
n
a
l
Pr
o
c
e
ss
in
g
,
v
o
l.
4
3
,
n
o
.
1
,
p
p
.
8
-
1
6
,
2
0
1
5
.
[5
]
N.
S
.
M
ish
ra
,
S
.
G
h
o
sh
a
n
d
A
.
G
h
o
sh
,
“
F
u
z
z
y
c
lu
ste
rin
g
a
lg
o
rit
h
m
s
in
c
o
rp
o
ra
ti
n
g
lo
c
a
l
i
n
f
o
rm
a
t
io
n
f
o
r
c
h
a
n
g
e
d
e
tec
ti
o
n
in
re
m
o
tely
s
e
n
se
d
im
a
g
e
s,”
Ap
p
li
e
d
S
o
ft
Co
m
p
u
ti
n
g
,
v
o
l.
1
2
,
n
o
.
8
,
p
p
.
2
6
8
3
-
2
6
9
2
,
2
0
1
2
.
[6
]
B.
N.
S
u
b
u
d
h
i,
F
.
Bo
v
o
lo
,
A
.
Gh
o
sh
a
n
d
L
.
Br
u
z
z
o
n
e
,
“
S
p
a
ti
o
-
c
o
n
tex
tu
a
l
f
u
z
z
y
c
lu
ste
rin
g
w
it
h
M
a
rk
o
v
ra
n
d
o
m
f
iel
d
m
o
d
e
l
f
o
r
c
h
a
n
g
e
d
e
tec
ti
o
n
in
re
m
o
tel
y
se
n
se
d
i
m
a
g
e
s,”
Op
ti
c
s
a
n
d
L
a
se
r
T
e
c
h
n
o
lo
g
y
,
v
o
l.
5
7
,
n
o
.
1
,
p
p
.
2
8
4
-
2
9
2
,
2
0
1
4
.
[7
]
C.
Be
n
e
d
e
k
,
M
.
S
h
a
d
a
y
d
e
h
,
Z.
Ka
to
,
T
.
S
z
irán
y
i
a
n
d
Z.
Zeru
b
ia
,
“
M
u
lt
il
a
y
e
r
M
a
rk
o
v
R
a
n
d
o
m
F
ield
m
o
d
e
ls
f
o
r
c
h
a
n
g
e
d
e
tec
ti
o
n
in
o
p
t
ica
l
re
m
o
te
se
n
sin
g
im
a
g
e
s,”
J
o
u
rn
a
l
o
f
Ph
o
t
o
g
r
a
mm
e
try
a
n
d
Rem
o
te
S
e
n
sin
g
,
v
o
l.
1
0
7
,
no.
1
,
p
p
.
2
2
-
3
7
,
2
0
1
5
.
[8
]
S
.
K.
Da
sh
a
n
d
M
.
P
a
n
d
a
,
“
Im
a
g
e
Clas
si
f
ic
a
ti
o
n
u
sin
g
Da
ta
M
in
in
g
Tec
h
n
iq
u
e
s,”
Ad
v
a
n
c
e
s
in
Co
mp
u
ter
S
c
ien
c
e
a
n
d
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
lo
g
y
(
ACS
IT
),
v
o
l
.
3
,
n
o.
3
,
p
p
.
1
5
7
-
1
6
2
,
2
0
1
6
.
[9
]
A
.
Bi
f
e
t,
R.
Kirk
b
y
,
G
.
Ho
l
m
e
s
a
n
d
B.
P
f
a
h
rin
g
e
r,
“
M
OA
:
M
a
ss
iv
e
On
li
n
e
A
n
a
l
y
sis,”
J
o
u
r
n
a
l
o
f
M
a
c
h
in
e
L
e
a
rn
in
g
Res
e
a
rc
h
,
v
o
l.
1
1
,
p
p
.
1
6
0
1
-
1
6
0
4
,
2
0
1
0
.
[1
0
]
R.
T
h
iru
m
a
h
a
l
a
n
d
P
.
A
.
De
e
p
a
l
i,
“
KN
N
a
n
d
A
R
L
Ba
se
d
I
m
p
u
tatio
n
t
o
Esti
m
a
te
M
issin
g
V
a
lu
e
s,”
In
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
En
g
in
e
e
rin
g
a
n
d
I
n
fo
rm
a
t
ics
,
v
o
l.
2
,
p
p
.
1
1
9
-
1
2
4
,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
r
o
p
o
s
ed
a
lg
o
r
ith
m
fo
r
ima
g
e
cla
s
s
i
fica
tio
n
u
s
in
g
r
eg
r
ess
io
n
-
b
a
s
ed
…
(
C
h
a
n
in
to
r
n
Jitta
w
ir
iya
n
u
ko
o
n
)
1027
[1
1
]
H.
Y.
M
u
ss
a
,
J.
B.
M
it
c
h
e
ll
a
n
d
R.
C.
G
len
,
“
F
u
ll
L
a
p
lac
ian
ise
d
P
o
ste
rio
r
Na
iv
e
Ba
y
e
sia
n
A
lg
o
rit
h
m
,
”
J
o
u
rn
a
l
o
f
Ch
e
min
fo
rm
a
ti
c
s
,
p
p
.
1
-
6
,
2
0
1
3
.
[1
2
]
W
.
Hu
,
W
.
Hu
a
n
d
S
.
M
a
y
b
a
n
k
,
“
A
d
a
Bo
o
st
-
Ba
se
d
A
lg
o
rit
h
m
f
o
r
Ne
t
w
o
rk
In
tru
sio
n
De
tec
ti
o
n
,
”
I
EE
E
T
ra
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
a
n
d
Cy
b
e
rn
e
ti
c
s
,
v
o
l.
3
8
,
n
o
.
2
,
p
p
.
5
7
7
-
5
8
2
,
2
0
0
8
.
[1
3
]
A
.
Uc
a
r,
Y.
De
m
ir
a
n
d
C.
G
u
z
e
li
s,
“
Ob
jec
t
re
c
o
g
n
it
i
o
n
a
n
d
d
e
tec
ti
o
n
w
it
h
d
e
e
p
lea
rn
i
n
g
f
o
r
a
u
t
o
n
o
m
o
u
s
d
riv
in
g
a
p
p
li
c
a
ti
o
n
s,”
T
ra
n
sa
c
ti
o
n
s
o
f
t
h
e
S
o
c
iety
fo
r
M
o
d
e
ll
in
g
a
n
d
S
im
u
la
ti
o
n
I
n
ter
n
a
t
io
n
a
l
,
v
o
l.
9
3
,
n
o
.
9
,
p
p
.
7
5
9
-
7
6
9
,
2
0
1
7
.
[1
4
]
K.
He
,
X
.
Z
h
a
n
g
,
S
.
Re
n
a
n
d
J.
S
u
n
,
“
De
e
p
Re
sid
u
a
l
L
e
a
rn
in
g
f
o
r
I
m
a
g
e
Re
c
o
g
n
it
io
n
,
”
Co
mp
u
ter
Vi
sio
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
p
p
.
1
-
1
2
,
2
0
1
5
.
[1
5
]
S
.
Krish
n
a
m
u
rth
y
a
n
d
R.
T
z
o
n
e
v
a
,
“
De
c
o
m
p
o
siti
o
n
-
Co
o
rd
i
n
a
ti
n
g
M
e
th
o
d
f
o
r
P
a
ra
ll
e
l
S
o
l
u
ti
o
n
o
f
a
M
u
lt
i
A
re
a
Co
m
b
in
e
d
Eco
n
o
m
ic
E
m
issio
n
Disp
a
tch
P
ro
b
lem
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
te
r
En
g
i
n
e
e
rin
g
,
v
o
l.
6
,
n
o
.
5
,
p
p
.
2
0
4
8
-
2
0
6
3
,
2
0
1
6
.
[1
6
]
S
.
F
.
C.
Ha
v
ian
a
a
n
d
M
.
T
a
u
f
ik
,
“
Co
m
p
a
riso
n
o
f
V
a
rio
u
s
S
im
il
a
rit
y
M
e
a
su
re
s
f
o
r
A
v
e
r
a
g
e
I
m
a
g
e
Ha
sh
in
M
o
b
i
le
P
h
o
n
e
A
p
p
li
c
a
ti
o
n
,
”
Pro
c
e
e
d
i
n
g
o
f
t
h
e
3
rd
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
El
e
c
trica
l
En
g
i
n
e
e
rin
g
,
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
In
f
o
rm
a
ti
c
s (
EE
CS
I)
,
v
o
l.
3
,
p
p
.
1
-
4
,
2
0
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.