TELK
O
M
N
I
K
A
Te
l
e
c
ommu
n
i
c
at
i
on
,
C
omp
u
ti
n
g
,
El
e
c
t
r
on
i
c
s
an
d
C
on
tr
o
l
V
ol
.
18
,
N
o
.
2
,
A
pri
l
2020
,
p
p.
69
5
~
70
4
IS
S
N
:
1693
-
6930
,
a
c
c
r
e
di
t
e
d
F
i
rs
t
G
r
a
de
by
K
e
m
e
n
ri
s
t
e
kdi
k
t
i
,
D
e
c
r
e
e
N
o:
21
/
E
/
K
P
T
/
2
018
D
O
I:
10
.
129
28/
T
E
L
K
O
M
N
IK
A
.
v1
8i
2
.
1372
6
695
Jou
r
n
al
h
o
m
e
pa
ge
:
ht
t
p:
/
/
j
our
nal
.
uad
.
a
c
.
i
d/
i
nd
e
x
.
php
/
T
E
L
K
O
MNIKA
N
e
w
a
p
p
r
o
a
c
h
t
o
t
h
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
o
f
t
h
e
e
a
s
y
e
x
p
r
e
ss
i
o
n
r
e
c
o
g
n
i
t
i
o
n
s
y
st
e
m
b
y
r
o
b
u
st
t
e
c
h
n
i
q
u
e
s
(
S
IFT
,
PC
A
-
S
IFT
,
A
S
IFT
a
n
d
S
U
R
F)
A
h
me
d
C
h
at
e
r
,
A
b
d
e
l
al
i
Las
far
L
a
bor
a
t
or
y
o
f
S
ys
t
e
m
A
na
l
ys
i
s
,
I
nf
o
r
m
a
t
i
o
n
P
r
o
c
e
s
s
i
ng
a
nd
i
ndu
s
t
r
y
M
a
n
a
ge
m
e
nt
,
H
i
gh
S
c
h
ool
o
f
T
e
c
hno
l
ogy
S
A
L
E
M
oha
m
m
a
di
a
E
ngi
ne
e
r
i
ng
S
c
h
o
ol
i
n
R
a
ba
t
,
M
or
o
c
c
o
A
r
ti
c
l
e
I
n
fo
A
BS
TR
A
C
T
Ar
t
i
c
l
e
h
i
s
t
or
y
:
Re
c
e
i
v
e
d
J
ul
25
,
20
19
Re
vi
s
e
d
J
a
n
16
,
202
0
A
c
c
e
pt
e
d
F
e
b
5
,
2020
I
n
r
e
c
e
n
t
ye
a
r
s
,
f
a
c
i
a
l
r
e
c
ogn
i
t
i
o
n
ha
s
be
e
n
a
m
a
j
o
r
pr
ob
l
e
m
i
n
t
he
f
i
e
l
d
of
c
o
m
p
ut
e
r
v
i
s
i
on
,
w
h
i
c
h
h
a
s
a
t
t
r
a
c
t
e
d
l
ot
s
of
i
n
t
e
r
e
s
t
i
n
pr
e
v
i
ous
ye
a
r
s
be
c
a
us
e
of
i
t
s
us
e
i
n
di
f
f
e
r
e
n
t
a
p
pl
i
c
a
t
i
on
s
by
di
f
f
e
r
e
nt
do
m
a
i
n
s
a
nd
i
m
a
ge
a
na
l
ys
i
s
.
W
h
i
c
h
i
s
ba
s
e
d
on
t
he
e
x
t
r
a
c
t
i
on
of
f
a
c
i
a
l
d
e
s
c
r
i
pt
or
s
,
i
t
i
s
a
v
e
r
y
i
m
por
t
a
nt
s
t
e
p
i
n
f
a
c
i
a
l
r
e
c
ogn
i
t
i
on
.
I
n
t
h
i
s
a
r
t
i
c
l
e
,
w
e
c
o
m
pa
r
e
d
r
ob
us
t
m
e
t
ho
ds
(
S
I
F
T
,
P
C
A
-
S
I
F
T
,
A
S
I
F
T
a
nd
S
U
R
F
)
t
o
e
x
t
r
a
c
t
r
e
l
e
v
a
nt
f
a
c
i
a
l
i
nf
or
m
a
t
i
on
w
i
t
h
d
i
f
f
e
r
e
n
t
f
a
c
i
a
l
po
s
t
ur
e
v
a
r
i
a
t
i
o
ns
(
op
e
n
a
nd
unope
ne
d
m
ou
t
h
,
gl
a
s
s
e
s
a
nd
no
g
l
a
s
s
e
s
,
ope
n
a
nd
c
l
o
s
e
d
e
ye
s
)
.
T
he
s
i
m
ul
a
t
i
on
r
e
s
ul
t
s
s
how
t
ha
t
t
he
de
t
e
c
t
o
r
(
S
U
R
F
)
i
s
be
t
t
e
r
t
h
a
n
o
t
he
r
s
a
t
f
i
nd
i
ng
t
he
s
i
m
i
l
a
r
i
t
y
de
s
c
r
i
pt
or
a
nd
c
a
l
c
u
l
a
t
i
on
t
i
m
e
.
O
u
r
m
e
t
hod
i
s
ba
s
e
d
on
t
he
no
r
m
a
l
i
z
a
t
i
o
n
of
ve
c
t
or
de
s
c
r
i
p
t
or
s
a
n
d
c
o
m
bi
n
e
d
w
i
t
h
t
h
e
R
A
N
S
A
C
a
l
go
r
i
t
h
m
t
o
c
a
nc
e
l
out
l
i
e
r
s
i
n
or
d
e
r
t
o
c
a
l
c
ul
a
t
e
t
he
H
e
s
s
i
a
n
m
a
t
r
i
x
w
i
t
h
t
h
e
ob
j
e
c
t
i
v
e
of
r
e
du
c
i
n
g
t
he
c
a
l
c
u
l
a
t
i
on
t
i
m
e
.
T
o
va
l
i
da
t
e
o
ur
e
xpe
r
i
e
n
c
e
,
w
e
t
e
s
t
e
d
f
ou
r
f
a
c
i
a
l
i
m
a
ge
s
da
t
a
b
a
s
e
s
c
o
nt
a
i
n
i
ng
s
e
v
e
r
a
l
m
o
di
f
i
c
a
t
i
o
ns
.
T
h
e
r
e
s
ul
t
s
of
t
he
s
i
m
ul
a
t
i
on
s
how
t
h
a
t
ou
r
m
e
t
ho
d
i
s
m
o
r
e
e
f
f
i
c
i
e
nt
t
ha
n
ot
he
r
de
t
e
c
t
o
r
s
i
n
t
e
r
m
s
of
s
p
e
e
d
o
f
r
e
c
ogn
i
t
i
on
a
n
d
d
e
t
e
r
m
i
na
t
i
on
of
s
i
m
i
l
a
r
poi
nt
s
be
t
w
e
e
n
t
w
o
i
m
a
ge
s
o
f
t
h
e
s
a
m
e
f
a
c
e
,
o
ne
b
e
l
o
ng
i
ng
t
o
t
he
b
a
s
e
of
t
h
e
t
e
xt
a
nd
t
he
o
t
he
r
one
t
o
t
he
b
a
s
e
d
r
i
ve
n
by
d
i
f
f
e
r
e
n
t
m
o
di
f
i
c
a
t
i
o
ns
.
T
hi
s
m
e
t
h
od,
w
hi
c
h
c
a
n
be
a
ppl
i
e
d
on
a
m
ob
i
l
e
p
l
a
t
f
or
m
t
o
a
na
l
yz
e
t
h
e
c
on
t
e
n
t
o
f
s
i
m
p
l
e
i
m
a
ge
s
,
f
or
e
xa
m
p
l
e
,
t
o
de
t
e
c
t
d
r
i
v
e
r
f
a
t
i
gue
,
hu
m
a
n
-
m
a
c
h
i
ne
i
nt
e
r
a
c
t
i
on
,
hum
a
n
-
r
ob
ot
.
U
s
i
n
g
de
s
c
r
i
pt
o
r
s
w
i
t
h
p
r
op
e
r
t
i
e
s
i
m
po
r
t
a
n
t
f
or
good
a
c
c
ur
a
c
y
a
nd
r
e
a
l
-
t
i
m
e
r
e
s
p
ons
e
.
Ke
y
w
or
d
s
:
F
a
c
e
r
e
c
og
ni
t
i
on
F
e
a
t
u
re
m
a
t
c
hi
n
g
F
our
d
a
t
a
ba
s
e
N
orm
a
l
i
z
a
t
i
on
of
fe
a
t
ure
RA
N
S
A
C
w
i
t
h
S
U
RF
Re
c
o
gni
t
i
on
r
a
t
e
T
hi
s
i
s
an
op
e
n
ac
c
e
s
s
ar
t
i
c
l
e
u
nde
r
t
he
C
C
B
Y
-
SA
l
i
c
e
n
s
e
.
Cor
r
e
s
pon
di
n
g
Au
t
h
or
:
A
hm
e
d
Ch
a
t
e
r,
L
a
bo
ra
t
ory
of
S
ys
t
e
m
A
n
a
l
ys
i
s
,
Inform
a
t
i
on
P
roc
e
s
s
i
ng
a
n
d
i
ndus
t
r
y
M
a
na
g
e
m
e
n
t
,
H
i
gh
S
c
hoo
l
of
T
e
c
hno
l
ogy
S
A
L
E
,
M
oha
m
m
e
d
V
U
ni
v
e
rs
i
t
y,
R
a
ba
t
,
M
oro
c
c
o.
E
m
a
i
l
:
a
h
m
e
d
c
h
a
t
e
r11
@
gm
a
i
l
.
c
o
m
,
a
l
i
.
l
a
s
fa
r
@
g
m
a
i
l
.
c
o
m
1.
I
N
TR
O
D
U
C
TI
O
N
E
a
s
y
r
e
c
ogni
t
i
o
n
h
a
s
be
e
n
a
f
i
e
l
d
of
re
s
e
a
r
c
h
i
nt
e
re
s
t
i
n
pre
vi
ous
y
e
a
rs
,
w
hi
c
h
i
s
us
e
d
by
di
ff
e
re
nt
a
ppl
i
c
a
t
i
ons
i
n
c
o
m
put
e
r
vi
s
i
o
n.
Curr
e
nt
l
y
a
dd
re
s
s
i
ng
t
h
e
pr
obl
e
m
s
t
ha
t
a
ri
s
e
fro
m
t
h
e
a
c
q
ui
s
i
t
i
on
of
i
m
a
g
e
s
w
i
t
h
di
ff
e
r
e
nt
va
ri
a
t
i
ons
i
n
pos
e
a
nd
v
a
ri
a
t
i
on
of
e
a
s
y
e
xpr
e
s
s
i
on.
T
h
e
re
a
re
s
om
e
t
r
a
di
t
i
on
a
l
a
l
gor
i
t
h
m
s
f
or
fa
c
e
r
e
c
ogni
t
i
o
n
s
u
c
h:
a
s
E
i
ge
nF
a
c
e
[1]
,
F
i
s
he
rF
a
c
e
[2
]
,
2D
-
P
CA
[3]
,
a
nd
E
l
a
s
t
i
c
G
r
a
ph
M
a
t
c
h
i
ng
[4]
.
T
he
S
c
a
l
e
Inv
a
ri
a
nt
F
e
a
t
ur
e
T
ra
ns
fo
rm
(S
IF
T
)
propos
e
d
by
D
a
vi
d
G
.
L
ow
e
[5
]
[
6]
,
h
a
s
be
e
n
w
i
de
l
y
us
e
d
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
69
5
-
7
04
696
obj
e
c
t
d
e
t
e
c
t
i
on
a
nd
re
c
ogn
i
t
i
on.
T
he
r
e
a
r
e
a
l
s
o
s
o
m
e
w
orks
on
t
h
e
us
e
of
S
IF
T
f
e
a
t
ure
s
i
n
f
a
c
e
re
c
ogn
i
t
i
on,
s
uc
h
a
s
S
IF
T
-
G
RID
prop
os
e
d
by
M
.
B
i
c
e
go
[7
]
a
nd
S
IF
T
CL
U
S
T
E
R
propos
e
d
by
J
un
L
uo
[
8].
M
a
ny
t
ra
d
i
t
i
ona
l
m
e
t
h
ods
c
a
n
e
xt
ra
c
t
t
h
e
re
m
a
rka
bl
e
poi
n
t
s
ba
s
e
d
on
l
oc
a
l
m
e
t
hods
,
a
nd
r
e
s
e
a
r
c
h
h
a
s
s
how
n
t
h
at
(S
IF
T
)
w
orks
w
e
l
l
i
n
fa
c
i
a
l
r
e
c
o
gni
t
i
on
,
b
ut
t
he
di
s
a
dva
n
t
a
ge
s
of
t
he
t
e
c
hn
i
qu
e
t
a
ke
a
l
ong
t
i
m
e
t
o
e
x
t
ra
c
t
t
he
d
e
s
c
r
i
pt
ors
a
nd
c
ha
ng
e
of
i
l
l
u
m
i
na
t
i
on
.
T
he
a
ut
ho
rs
propos
e
o
t
he
r
m
e
t
hods
w
h
i
c
h
i
s
a
l
r
e
a
dy
us
e
d
t
o
re
d
uc
e
c
o
m
pu
t
a
t
i
on
t
i
m
e
fo
r
e
x
a
m
p
l
e
:
K
d
-
t
r
e
e
i
s
us
e
d
i
n
t
h
e
s
e
a
r
c
h
s
t
e
p
for
t
he
ne
a
re
s
t
c
onn
e
c
t
i
o
n,
a
nd
P
CA
i
s
prop
os
e
d
t
o
re
duc
e
t
he
d
i
m
e
ns
i
o
ns
of
S
IF
T
c
h
a
r
a
c
t
e
ri
s
t
i
c
s
i
s
c
a
l
l
e
d
P
CA
-
S
IF
T
.
H
ow
e
ve
r
,
t
h
e
t
e
c
h
ni
qu
e
S
IF
T
s
t
i
l
l
do
no
t
a
l
l
ow
t
o
re
s
pon
d
t
he
r
e
qu
i
re
m
e
n
t
s
of
on
l
i
n
e
s
e
rv
i
c
e
s
.
T
he
n
,
t
h
e
a
u
t
hors
propos
e
a
no
t
he
r
d
e
t
e
c
t
or
of
t
h
e
s
a
m
e
p
e
rfo
r
m
a
n
c
e
a
s
S
IF
T
,
c
a
l
l
e
d
(S
U
RF
)
i
s
a
de
t
e
c
t
or
ro
bus
t
i
n
t
e
r
m
of
di
ffe
r
e
nt
t
ra
ns
f
orm
a
t
i
on
of
fa
c
e
l
i
k
e
w
e
m
e
nt
i
on
e
d
i
n
t
he
four
t
h
t
i
t
l
e
.
F
i
rs
t
l
y
(S
U
RF
)
i
s
us
e
d
t
o
e
x
t
r
a
c
t
t
he
r
e
m
a
rk
a
bl
e
po
i
nt
s
us
i
ng
m
a
t
ri
x
a
p
p
roxi
m
a
t
i
on
of
H
e
s
s
i
a
n
a
ppl
i
c
a
t
e
on
t
he
i
n
t
e
gr
a
l
i
m
a
g
e
s
i
n
orde
r
t
o
l
oc
a
t
e
de
s
c
r
i
pt
or
t
ha
t
a
l
l
ow
t
o
d
e
c
r
e
a
s
e
t
he
a
na
l
ys
i
s
t
i
m
e
of
i
m
a
g
e
,
a
nd
t
he
n
w
e
us
e
t
h
e
w
a
v
e
l
e
t
s
i
n
t
he
x
a
n
d
y
di
re
c
t
i
ons
t
o
d
e
s
c
r
i
be
t
he
d
i
s
t
ri
but
i
on
of
t
he
i
nt
e
ns
i
t
y
i
n
t
h
e
vi
c
i
ni
t
y
of
t
he
r
e
m
a
rk
a
bl
e
poi
nt
s
,
i
n
a
dd
i
t
i
on
t
he
de
t
e
c
t
o
r
(S
U
RF
)
us
e
d
on
l
y
(64)
di
m
e
ns
i
on
t
o
d
e
c
r
e
a
s
e
t
h
e
t
i
m
e
of
c
a
l
c
ul
a
t
i
on.
In
t
hi
s
a
rt
i
c
l
e
,
w
e
ha
ve
pr
e
s
e
n
t
e
d
t
he
c
o
m
pa
r
i
s
on
be
t
w
e
e
n
t
h
e
di
ff
e
re
nt
rob
us
t
de
t
e
c
t
o
rs
t
e
s
t
e
d
i
n
our
pre
vi
ous
w
ork
by
d
i
ff
e
re
n
t
va
ri
a
t
i
o
ns
i
n
v
i
e
w
po
i
nt
s
[
9
,
10
]
.
O
ur
m
e
t
hod
b
a
s
e
d
o
n
t
h
e
de
t
e
c
t
o
r
(S
U
RF
)
w
i
t
h
t
he
RA
N
S
A
C
[11
]
a
l
gor
i
t
h
m
w
hi
c
h
e
s
t
i
m
a
t
e
s
t
h
e
d
a
t
a
b
e
t
w
e
e
n
t
hr
e
e
s
t
e
ps
.
T
h
e
f
i
rs
t
one
c
ons
i
s
t
s
t
o
e
x
t
ra
c
t
de
s
c
r
i
pt
ors
.
T
h
e
s
e
c
ond
on
e
c
ho
os
e
s
ra
nd
om
e
nt
ry
p
oi
nt
s
a
nd
t
h
e
n
e
s
t
i
m
a
t
e
s
t
h
e
s
e
p
a
ra
m
e
t
e
rs
by
t
he
a
dj
us
t
m
e
nt
m
odu
l
e
.
T
h
e
t
h
i
rd
s
t
e
p
c
o
m
p
a
re
s
t
h
e
s
e
p
a
r
a
m
e
t
e
rs
by
t
h
e
c
o
m
pa
t
i
b
i
l
i
t
y
o
f
t
h
e
a
dj
us
t
m
e
n
t
m
odu
l
e
b
a
s
e
d
on
a
c
e
rt
a
i
n
m
a
xi
m
u
m
e
rror
t
hr
e
s
hol
d
for
ob
j
e
c
t
i
v
e
t
o
c
a
n
c
e
l
ou
t
ou
t
l
i
e
rs
a
nd
no
i
s
e
,
l
ook
a
t
t
he
a
l
g
ori
t
hm
be
l
ow
.
T
h
e
r
e
s
ul
t
of
t
he
c
o
m
pa
r
i
s
on
of
our
m
e
t
hod
g
i
v
e
s
a
g
ood
r
e
s
ul
t
i
n
t
e
rm
s
of
s
pe
e
d
of
c
orre
s
p
onde
nc
e
c
a
l
c
u
l
a
t
i
on
a
nd
re
c
ogn
i
t
i
on
r
a
t
e
w
i
t
h
d
i
ffe
r
e
nt
v
a
ri
a
t
i
ons
of
f
a
c
e
c
h
a
ng
e
o
f
t
he
s
a
m
e
pe
rs
on.
2.
D
ES
C
R
I
P
TO
R
E
X
TR
A
C
TI
O
N
BY
R
O
BU
S
T
M
ET
H
O
D
S
K
e
y
p
oi
n
t
e
xt
ra
c
t
i
on
t
e
c
hn
i
qu
e
s
a
re
b
a
s
e
d
on
i
n
va
r
i
a
n
t
t
o
a
ffi
n
e
t
r
a
ns
for
m
a
t
i
ons
a
m
ong
t
he
s
e
t
e
c
hni
q
ue
s
(S
IF
T
,
A
S
IF
T
,
P
CA
-
S
IF
T
a
nd
S
U
RF
)
.
T
he
qua
l
i
t
y
of
op
e
ra
t
i
on
of
a
f
a
c
i
a
l
re
c
ogn
i
t
i
on
s
ys
t
e
m
i
s
l
i
nk
e
d
t
o
t
he
c
hoi
c
e
of
d
e
t
e
c
t
or
for
f
e
a
t
ur
e
e
xt
r
a
c
t
i
o
n
b
e
c
a
u
s
e
e
a
c
h
t
e
c
h
ni
qu
e
i
s
a
da
p
t
e
d
t
o
a
gi
ve
n
c
on
t
e
xt
.
W
e
c
hos
e
t
he
S
pe
e
d
e
d
-
U
p
Robus
t
F
e
a
t
ur
e
s
(S
U
RF
)
d
e
t
e
c
t
or
ove
r
ot
he
r
m
e
t
hods
be
c
a
us
e
of
i
t
s
ro
bus
t
n
e
s
s
a
nd
t
he
us
e
of
s
e
c
ond
-
ord
e
r
G
a
us
s
i
a
n
p
a
rt
i
a
l
d
e
ri
v
a
t
i
ve
s
,
w
hi
c
h
i
m
p
rove
t
h
e
t
i
m
e
of
r
e
a
l
-
t
i
m
e
i
m
a
ge
a
na
l
ys
i
s
.
Th
e
d
i
ffe
re
n
t
s
t
e
ps
of
t
he
a
l
gor
i
t
h
m
(S
U
RF
)
for
t
he
e
xt
r
a
c
t
i
on
of
t
he
k
e
y
po
i
nt
s
fo
l
l
ow
t
he
fo
l
l
ow
i
ng
s
t
e
ps
(H
e
s
s
i
a
n
m
a
t
r
i
x
-
b
a
s
e
d
i
nt
e
re
s
t
po
i
nt
s
,
I
nt
e
re
s
t
po
i
nt
de
s
c
ri
pt
i
on
a
nd
de
s
c
r
i
pt
or
c
o
m
pon
e
nt
s
)
.
2.
1
.
Th
e
or
y
of
s
u
r
f
(s
p
e
e
d
e
d
-
u
p
r
ob
u
s
t
fe
atu
r
e
s
)
In
200
6,
B
a
y
e
t
a
l
.
[12
]
pro
pos
e
a
n
e
w
m
e
t
hod
of
l
o
c
a
l
de
s
c
ri
pt
i
on
of
po
i
nt
s
of
i
n
t
e
r
e
s
t
.
N
a
m
e
d
S
U
RF
(S
pe
e
d
-
U
p
Robus
t
F
e
a
t
ure
s
)
.
S
t
ro
ngl
y
i
nfl
ue
n
c
e
d
by
t
h
e
S
I
F
T
a
ppro
a
c
h,
i
t
c
oupl
e
s
a
s
t
e
p
of
r
e
gi
s
t
ra
t
i
on
of
t
he
a
n
a
l
ys
i
s
a
r
e
a
w
i
t
h
t
he
c
ons
t
ruc
t
i
on
of
a
h
i
s
t
og
ra
m
of
ori
e
n
t
e
d
gra
d
i
e
nt
s
.
T
h
e
c
a
l
c
u
l
a
t
i
o
n
pro
c
e
s
s
c
ons
i
s
t
s
i
n
de
t
e
rm
i
ni
ng
t
h
e
r
ot
a
t
i
on
(or
re
c
ord
i
ng)
a
n
gl
e
t
o
b
e
a
ppl
i
e
d
t
o
t
he
l
o
c
a
l
de
s
c
ri
p
t
i
on
w
i
ndow
.
T
o
t
h
i
s
,
t
h
e
a
u
t
hors
a
ppl
y
H
a
a
r
w
a
ve
l
e
t
s
t
o
t
he
i
n
t
e
gr
a
l
i
m
a
g
e
,
t
hus
s
i
gni
f
i
c
a
nt
l
y
r
e
duc
i
ng
c
o
m
put
a
t
i
on
t
i
m
e
.
T
he
s
e
w
a
v
e
l
e
t
s
m
a
ke
i
t
p
os
s
i
bl
e
t
o
c
a
l
c
u
l
a
t
e
t
h
e
f
i
rs
t
de
r
i
va
t
i
v
e
s
of
t
h
e
i
m
a
g
e
on
a
s
qu
a
re
n
e
i
ghb
orhoo
d
a
nd
t
hus
t
o
s
t
udy
t
he
d
i
s
t
ri
b
ut
i
on
of
t
h
e
hor
i
z
on
t
a
l
a
n
d
ve
r
t
i
c
a
l
g
ra
d
i
e
n
t
s
.
T
he
r
e
s
pons
e
s
of
t
h
e
w
a
v
e
l
e
t
s
t
h
e
n
m
a
k
e
i
t
p
os
s
i
bl
e
t
o
dra
w
t
h
e
gr
a
ph
of
d
i
s
t
ri
b
ut
i
on
of
t
h
e
gr
a
di
e
nt
s
a
nd
t
o
d
e
du
c
e
t
he
r
e
fro
m
t
he
a
ng
l
e
of
re
gi
s
t
r
a
t
i
on
.
O
n
t
he
i
ni
t
i
a
l
i
m
a
ge
t
he
c
i
rc
l
e
r
e
pr
e
s
e
n
t
s
t
h
e
re
gi
on
of
i
nt
e
re
s
t
w
hos
e
ra
di
us
i
s
e
q
ua
l
t
o
6s
w
he
r
e
s
c
o
rre
s
po
nds
t
o
t
he
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
c
a
l
e
e
x
t
ra
c
t
e
d
fro
m
t
he
fa
s
t
-
H
e
s
s
i
a
n
de
t
e
c
t
or
2.
2
.
H
e
s
s
i
an
matr
i
x
-
b
as
e
d
i
n
t
e
r
e
s
t
p
o
i
n
ts
T
he
S
U
RF
d
e
t
e
c
t
or
i
s
ba
s
e
d
on
t
he
d
e
t
e
r
m
i
na
nt
of
t
he
H
e
s
s
i
a
n
m
a
t
r
i
x
[12
]
.
In
or
de
r
t
o
m
o
t
i
v
a
t
e
t
h
e
us
e
of
t
he
H
e
s
s
i
a
n,
w
e
c
ons
i
d
e
r
a
c
o
nt
i
nuous
func
t
i
o
n
o
f
t
w
o
va
ri
a
bl
e
s
s
u
c
h
t
ha
t
t
h
e
v
a
l
u
e
of
t
h
e
f
unc
t
i
on
a
t
(
,
)
xy
i
s
g
i
ve
n
by
(
,
)
f
x
y
.
T
h
e
H
e
s
s
i
a
n
m
a
t
r
i
x
(H
)
i
s
t
he
m
a
t
ri
x
of
p
a
r
t
i
a
l
d
e
ri
va
t
e
of
t
he
fun
c
t
i
on
(
,
)
f
x
y
.
W
he
re
(1)
T
he
de
t
e
rm
i
na
n
t
of
t
hi
s
m
a
t
r
i
x
,
C
a
l
l
e
d
d
i
s
c
ri
m
i
na
t
i
on
,
i
s
c
a
l
c
ul
a
t
e
d
a
s
fol
l
ow
s
:
22
2
22
2
(
(
,
)
)
ff
x
x
y
H
f
x
y
ff
x
y
y
=
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
Ne
w
appr
oa
c
h
t
o
t
h
e
i
de
nt
i
f
i
c
at
i
on
of
t
h
e
e
as
y
e
x
pr
e
s
s
i
on
r
e
c
ogni
t
i
on
s
y
s
t
e
m
…
(
A
hm
e
d
C
hat
e
r
)
697
2
2
2
2
d
e
t(
)
22
f
f
f
H
xy
xy
=−
(
2
)
T
he
di
s
c
ri
m
i
na
t
i
ng
v
a
l
u
e
i
s
us
e
d
t
o
d
e
t
e
rm
i
ne
t
he
m
a
xi
m
um
a
nd
m
i
n
i
m
u
m
of
(2)
.
I
f
t
he
r
e
s
ul
t
of
t
he
produ
c
t
of
t
h
e
ne
g
a
t
i
ve
e
i
ge
nv
a
l
u
e
s
t
h
e
po
i
nt
s
i
s
n
ot
a
l
o
c
a
l
e
x
t
re
m
a
,
t
he
n
i
f
t
h
e
pro
duc
t
s
of
t
he
p
os
i
t
i
ve
e
i
g
e
nv
a
l
u
e
s
va
l
ue
t
he
po
i
nt
s
c
l
a
s
s
i
fi
e
d
a
s
e
x
t
r
e
m
a
.
In
[6]
a
nd
[8]
i
t
d
e
s
c
r
i
be
s
t
h
a
t
a
H
e
s
s
i
a
n
m
a
t
ri
x
c
a
n
be
do
ne
a
s
a
gr
e
a
t
de
t
e
c
t
o
r
for
i
t
s
hi
g
h
produ
c
t
i
on
i
n
c
o
m
pu
t
a
t
i
on
a
l
t
i
m
e
a
nd
pr
e
c
i
s
i
on
.
S
c
a
l
e
ra
nge
c
a
n
be
obt
a
i
n
e
d
t
hroug
h
t
he
de
t
e
r
m
i
n
a
nt
of
t
he
H
e
s
s
i
a
n
or
H
e
s
s
i
a
n
–
L
a
pl
a
c
e
d
e
t
e
c
t
or.
G
i
v
e
n
a
po
i
n
t
(
)
,
P
x
y
i
n
t
he
i
m
a
ge
I,
t
he
H
e
s
s
i
a
n
m
a
t
ri
x
H
(p
,
σ
)
i
n
p
a
t
s
c
a
l
e
σ
i
s
d
e
fi
n
e
d
a
s
fol
l
ow
s
(3)
:
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
x
x
x
y
y
x
y
y
L
P
L
P
HX
L
P
L
P
=
(3)
W
he
r
e
2
()
(
,
)
(
)
*
2
g
L
P
I
x
xx
x
=
.
T
h
e
c
onvo
l
ut
i
on
of
t
h
e
s
e
c
ond
ord
e
r
G
a
us
s
i
a
n
d
e
r
i
va
t
i
ve
2
2
()
g
x
w
i
t
h
t
h
e
i
m
a
ge
a
t
p
oi
n
t
()
IP
a
nd
s
i
m
i
l
a
rl
y
by
yy
L
,
a
nd
2
()
(
,
)
(
)
*
xy
g
L
P
I
P
xy
=
.
T
he
s
e
de
r
i
va
t
i
v
e
s
a
r
e
know
n
a
s
L
a
pl
a
c
i
a
n
of
G
a
us
s
i
a
ns
.
B
a
s
e
d
on
t
he
t
urn
i
ndi
c
a
t
or
,
w
e
c
a
n
c
om
p
ut
e
t
he
de
t
e
r
m
i
na
n
t
of
t
he
H
e
s
s
i
a
n
f
or
e
a
c
h
pi
x
e
l
i
n
t
h
e
i
m
a
g
e
a
nd
us
e
t
he
pow
e
r
t
o
a
nd
t
h
e
re
m
a
r
ka
b
l
e
poi
nt
s
.
T
he
n
t
h
e
h
e
s
s
i
a
n
de
t
e
rm
i
na
n
t
c
a
l
c
u
l
a
t
e
s
t
o
e
xt
r
a
c
t
t
he
re
m
a
r
ka
bl
e
.
L
ow
e
[4
]
foun
d
a
pe
rfor
m
a
n
c
e
i
n
c
re
a
s
e
i
n
a
pprox
i
m
a
t
i
ng
t
h
e
L
a
pl
a
c
i
a
n
of
G
a
us
s
i
a
ns
by
a
d
i
ff
e
re
n
c
e
of
G
a
us
s
i
a
ns
.
In
a
s
i
m
i
l
a
r
m
a
nn
e
r,
Ba
y
[
13
]
propos
e
d
a
n
a
pp
roxi
m
a
t
i
o
n
t
o
t
he
L
a
p
l
a
c
i
a
n
o
f
G
a
us
s
i
a
ns
b
y
us
i
ng
box
-
l
t
e
r
r
e
pr
e
s
e
nt
a
t
i
ons
of
t
he
re
s
p
e
c
t
i
v
e
ke
rne
l
s
.
T
he
(S
U
RF
)
a
ppr
oa
c
h
e
xc
e
e
ds
(S
IF
T
)
i
n
t
e
r
m
s
of
s
p
e
e
d
t
o
c
a
l
c
u
l
a
t
e
p
oi
n
t
s
of
i
n
t
e
r
e
s
t
a
nd
t
h
e
i
r
a
c
c
u
ra
c
y.
S
U
RF
us
e
s
t
h
e
bu
i
l
t
-
i
n
i
m
a
ge
b
ox
fi
l
t
e
r
a
ga
i
ns
t
t
h
e
(S
IF
T
)
a
ppro
a
c
h
t
o
a
ppl
y
t
h
e
fi
l
t
e
r
t
o
e
a
c
h
i
m
a
ge
s
i
z
e
i
n
t
h
e
i
m
a
g
e
pyra
m
i
d.
In
t
h
e
S
U
RF
s
t
ra
t
e
gy
,
I
n
t
he
S
U
RF
a
ppro
a
c
h,
t
h
e
box
fi
l
t
e
r
i
n
(F
i
g
ure
1)
s
t
a
rt
s
w
i
t
h
a
9×
9
s
i
z
e
f
i
l
t
e
r
a
s
t
he
i
n
i
t
i
a
l
s
c
a
l
e
l
a
y
e
r
w
he
r
e
i
t
i
s
r
e
fe
r
re
d
t
o
a
s
s
c
a
l
e
s
=
1
.
2
(t
he
a
ppr
oxi
m
a
t
e
d
G
a
us
s
i
a
n
d
e
ri
v
a
t
i
ve
w
i
t
h
t
h
e
v
a
l
u
e
σ=
1.
2
)
a
n
d
i
ns
t
e
a
d
of
ha
v
i
ng
i
m
a
ge
pyra
m
i
ds
,
t
he
ori
gi
n
a
l
i
m
a
ge
w
i
l
l
be
fi
l
t
e
re
d
b
y
b
i
gg
e
r
m
a
s
ks
,
de
n
ot
e
d
t
h
e
m
by
xx
D
,
xy
D
,
a
nd
yy
D
.
H
e
s
s
i
a
n
d
e
t
e
rm
i
na
nt
us
i
ng
t
he
a
pprox
i
m
a
t
e
d
G
a
us
s
i
a
ns
a
nd
i
t
i
s
e
xpre
s
s
e
d
a
s
fo
l
l
o
w
s
(4):
2
d
e
t
(
)
(
)
a
p
p
r
o
x
x
x
y
y
x
y
H
D
D
w
D
=−
(
4
)
T
he
r
e
l
a
t
i
ve
w
e
i
g
ht
e
qua
l
s
0
.
9
w
=
,
t
o
ba
l
a
n
c
e
t
h
e
e
xp
re
s
s
i
on
of
(4)
.
W
h
i
c
h
a
l
l
ow
s
t
he
c
o
ns
e
rv
a
t
i
on
of
e
ne
rgy
be
t
w
e
e
n
G
a
us
s
i
a
n
n
uc
l
e
i
a
pprox
i
m
a
t
e
d
a
n
d
G
a
us
s
i
a
n
nu
c
l
e
i
.
F
i
gure
1
.
L
a
p
l
a
c
i
a
n
of
G
a
us
s
i
a
n
A
p
proxi
m
a
t
i
on.
T
op
Row
:
T
he
d
i
s
c
r
e
t
i
z
e
d
a
nd
c
r
opp
e
d
s
e
c
on
d
or
de
r
G
a
us
s
i
a
n
d
e
ri
v
a
t
i
ve
s
i
n
t
h
e
,
xy
a
nd
xy
-
di
re
c
t
i
ons
.
W
e
re
f
e
r
t
o
t
h
e
s
e
a
s
L
xx
,
L
xx
,
a
n
d
L
xy
.
2.
3
.
I
n
te
r
e
s
t
p
oi
n
t
d
e
s
c
r
i
p
t
i
on
an
d
d
e
s
c
r
i
p
to
r
c
o
mp
on
e
n
ts
.
H
a
a
r
w
a
ve
l
e
t
t
ha
t
a
l
l
ow
s
e
x
t
ra
c
t
i
ng
t
he
prop
e
rt
i
e
s
o
f
t
he
po
i
nt
s
of
i
nt
e
re
s
t
,
t
he
n
de
t
e
rm
i
ne
t
he
or
i
e
n
t
a
t
i
o
n
i
n
bo
t
h
t
h
e
(x
a
nd
y)
d
i
re
c
t
i
on
a
s
fo
l
l
ow
s
.
F
i
r
s
t
,
c
r
e
a
t
e
a
s
qua
r
e
re
gi
on
l
oc
a
l
i
z
e
d
on
t
he
p
oi
n
t
s
of
i
n
t
e
re
s
t
,
a
nd
t
he
n
de
t
e
rm
i
ne
t
h
e
di
ff
e
r
e
nt
di
r
e
c
t
i
o
n
t
ha
t
i
nt
r
oduc
e
d
i
n
[14
]
.
S
e
c
on
d,
w
e
di
vi
d
e
t
he
m
a
i
n
r
e
g
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
69
5
-
7
04
698
i
nt
o
a
n
e
qu
a
l
s
ub
-
r
e
g
i
on
(
4x4
)
a
s
s
how
n
i
n
F
i
gur
e
2,
w
h
i
c
h
pre
s
e
rv
e
s
t
h
e
da
t
a
of
i
n
t
e
re
s
t
for
e
a
c
h
s
ub
-
r
e
gi
o
n.
T
he
n
w
e
d
e
t
e
rm
i
ne
t
he
w
a
v
e
l
e
t
re
s
p
ons
e
s
of
H
a
a
r
a
t
5x5
by
e
qui
d
i
s
t
a
nc
e
poi
n
t
s
.
W
i
t
h
dx:
w
a
ve
l
e
t
r
e
s
pons
e
i
n
t
h
e
hori
z
on
t
a
l
d
i
r
e
c
t
i
on
a
nd
dy:
w
a
ve
l
e
t
r
e
s
pons
e
i
n
t
h
e
ve
r
t
i
c
a
l
d
i
r
e
c
t
i
on
(fi
l
t
e
r
s
i
z
e
2s
)
.
T
o
i
nc
r
e
a
s
e
t
he
robus
t
ne
s
s
t
ow
a
rds
ge
om
e
t
ri
c
d
e
for
m
a
t
i
o
ns
a
n
d
l
o
c
a
l
i
z
a
t
i
o
n
e
rrors
,
t
he
r
e
s
pons
e
s
dx
a
nd
dy
a
re
f
i
rs
t
w
e
i
g
ht
e
d
w
i
t
h
a
G
a
us
s
i
a
n
3
.
3
s
=
c
e
nt
e
r
e
d
a
t
t
he
i
n
t
e
r
e
s
t
i
n
g
po
i
nt
.
T
hi
r
d,
be
for
e
prov
i
di
ng
d
a
t
a
on
t
he
p
ol
a
ri
t
y
of
t
he
i
nt
e
ns
i
t
y
v
a
ri
a
t
i
ons
,
w
e
a
l
s
o
de
t
e
r
m
i
n
e
d
t
he
s
u
m
of
t
h
e
a
bs
o
l
ut
e
va
l
u
e
s
of
t
h
e
re
s
p
ons
e
s
,
dx
a
nd
dy
.
S
o,
e
a
c
h
s
u
b
-
re
gi
on
ha
s
a
fo
ur
-
di
m
e
ns
i
on
a
l
de
s
c
ri
p
t
or
v
e
c
t
or
(v)
fo
r
i
t
s
unde
r
l
yi
n
g
i
nt
e
ns
i
t
y
c
ons
t
r
uc
t
i
on
a
s
fol
l
ow
s
:
(
)
,
,
,
x
y
x
y
v
d
d
d
d
=
(5)
By
c
on
c
a
t
e
n
a
t
i
ng
a
l
l
s
ub
-
re
g
i
ons
of
d
i
m
e
ns
i
on
(
4x
4)
,
t
o
o
bt
a
i
n
a
d
e
s
c
r
i
pt
or
ve
c
t
or
o
f
l
e
ngt
h
6
4.
T
he
w
a
v
e
l
e
t
re
s
pons
e
s
a
r
e
i
nv
a
r
i
a
nt
i
n
i
l
l
u
m
i
n
a
t
i
on
.
Inv
a
r
i
a
nc
e
t
o
c
ont
r
a
s
t
(a
s
c
a
l
e
fa
c
t
o
r)
i
s
obt
a
i
ne
d
by
c
onv
e
rt
i
ng
t
h
e
d
e
s
c
r
i
pt
o
r
i
nt
o
a
un
i
t
ve
c
t
or.
F
i
g
ure
2.
C
ons
t
ru
c
t
i
on
t
h
e
4
-
di
m
e
ns
i
ona
l
de
s
c
ri
p
t
or
(S
U
RF
)
3.
D
A
TA
BA
S
E
AND
A
LG
O
R
I
TH
M
P
R
O
P
O
S
ED
O
ne
of
t
he
m
os
t
i
m
por
t
a
n
t
a
s
pe
c
t
s
of
t
h
e
d
e
ve
l
op
m
e
n
t
of
ne
w
re
c
ogni
t
i
o
n
s
ys
t
e
m
o
r
fa
c
i
a
l
e
xpr
e
s
s
i
on
de
t
e
c
t
i
on
i
s
t
h
e
c
ho
i
c
e
o
f
t
he
d
a
t
a
ba
s
e
t
ha
t
w
i
l
l
b
e
us
e
d
t
o
t
e
s
t
t
hi
s
s
ys
t
e
m
.
In
a
dd
i
t
i
on
,
c
o
m
m
on
d
a
t
a
ba
s
e
s
a
re
n
e
c
e
s
s
a
ry
t
o
be
n
c
h
m
a
rk
our
a
p
proa
c
h
.
In
t
h
i
s
a
r
t
i
c
l
e
,
w
e
w
i
l
l
pre
s
e
nt
fo
ur
(O
RL
,
G
ri
m
a
c
e
,
F
a
c
e
s
95
a
nd
F
a
c
e
s
96
)
.
P
opu
l
a
r
e
a
s
y
e
xp
re
s
s
i
on
da
t
a
b
a
s
e
s
t
h
a
t
a
r
e
pu
bl
i
c
l
y
a
nd
fr
e
e
l
y
a
va
i
l
a
bl
e
t
o
b
e
us
e
d
t
o
e
va
l
u
a
t
e
our
a
l
gor
i
t
h
m
.
3
.1
.
D
atab
as
e
O
R
L
T
he
a
l
gori
t
h
m
s
w
e
r
e
e
va
l
ua
t
e
d
t
o
v
a
l
i
da
t
e
our
e
xp
e
ri
e
nc
e
,
w
e
us
e
d
d
a
t
a
ba
s
e
a
re
f
e
r
e
nc
e
O
R
L
i
n
[1
5]
w
hi
c
h
c
ont
a
i
ns
40
s
ubj
e
c
t
s
a
nd
e
a
c
h
s
u
bj
e
c
t
c
on
t
a
i
ns
10
d
i
ff
e
re
n
t
fa
c
e
s
of
t
h
e
s
a
m
e
pe
rs
on.
T
he
r
e
s
ol
u
t
i
on
of
a
l
l
8
-
bi
t
i
m
a
g
e
s
i
s
11
2×
92
.
P
i
x
e
l
,
w
i
t
h
25
6
gra
ys
c
a
l
e
s
p
e
r
pi
xe
l
,
t
he
di
ff
e
re
nt
v
a
ri
a
t
i
o
ns
t
h
a
t
a
r
e
:
c
ha
n
ge
of
l
i
gh
t
i
ng
a
nd
va
ri
a
t
i
on
of
e
a
s
y
e
xpr
e
s
s
i
on
(op
e
n
m
out
h
a
nd
not
ope
n,
g
l
a
s
s
e
s
a
n
d
n
ot
g
l
a
s
s
e
s
,
e
ye
s
op
e
n
a
n
d
c
l
os
e
d).
A
s
s
how
n
i
n
F
i
gure
3.
F
i
g
ure
3
.
E
xa
m
p
l
e
of
i
m
a
g
e
s
of
f
a
c
e
s
of
t
w
o
i
ndi
v
i
du
a
l
s
fro
m
t
he
d
a
t
a
ba
s
e
(O
R
L
)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
Ne
w
appr
oa
c
h
t
o
t
h
e
i
de
nt
i
f
i
c
at
i
on
of
t
h
e
e
as
y
e
x
pr
e
s
s
i
on
r
e
c
ogni
t
i
on
s
y
s
t
e
m
…
(
A
hm
e
d
C
hat
e
r
)
699
3
.
2
.
D
at
ab
as
e
G
r
i
mac
e
T
he
a
l
gor
i
t
h
m
i
s
e
v
a
l
u
a
t
e
d
t
o
v
a
l
i
da
t
e
ou
r
e
xp
e
ri
e
nc
e
,
w
e
us
e
d
da
t
a
b
a
s
e
a
r
e
fe
r
e
n
c
e
G
ri
m
a
c
e
i
n
[16
]
w
hi
c
h
c
ont
a
i
ns
18
s
ubj
e
c
t
s
a
nd
e
a
c
h
s
u
bj
e
c
t
c
on
t
a
i
ns
20
d
i
ff
e
re
n
t
fa
c
e
s
of
t
h
e
s
a
m
e
pe
rs
on.
T
he
r
e
s
ol
u
t
i
on
of
a
l
l
8
-
bi
t
i
m
a
ge
s
i
s
180x
200
p
i
xe
l
,
w
i
t
h
256
gra
ys
c
a
l
e
s
pe
r
pi
x
e
l
,
t
h
e
d
i
ffe
re
n
t
v
a
ri
a
t
i
ons
t
h
a
t
a
r
e
:
(Ba
c
kgro
unds
,
H
e
a
d
S
c
a
l
e
,
H
e
a
d
t
urn
,
t
i
l
t
a
nd
s
l
a
nt
,
P
os
i
t
i
on
of
t
h
e
fa
c
e
i
n
t
he
i
m
a
g
e
,
Im
a
g
e
l
i
ght
i
ng
va
r
i
a
t
i
on
,
E
xpr
e
s
s
i
on
V
a
ri
a
t
i
on,
A
dd
i
t
i
ona
l
c
o
m
m
e
nt
)
.
A
s
s
how
n
i
n
F
i
gur
e
4
.
F
ig
ure
4.
E
xa
m
p
l
e
of
f
a
c
e
i
m
a
ge
s
of
t
w
o
i
nd
i
vi
dua
l
s
fr
om
t
h
e
G
ri
m
a
c
e
d
a
t
a
s
e
t
3
.
3
.
D
atab
as
e
F
ac
e
s
95
W
e
us
e
d
t
he
F
a
c
e
s
9
5
re
f
e
r
e
nc
e
d
a
t
a
ba
s
e
i
n
[1
7
]
w
hi
c
h
c
ont
a
i
ns
72
s
u
bj
e
c
t
s
a
nd
e
a
c
h
s
ubj
e
c
t
c
ont
a
i
ns
20
di
ff
e
re
nt
f
a
c
e
s
of
t
h
e
s
a
m
e
pe
rs
on
.
T
he
r
e
s
ol
u
t
i
on
of
a
l
l
8
-
bi
t
i
m
a
g
e
s
i
s
180x
200,
t
h
e
di
ff
e
r
e
nt
v
a
r
i
a
t
i
ons
be
i
ng
:
(
Ba
c
kgrou
nds
:
H
e
a
d
S
c
a
l
e
,
H
e
a
d
t
urn,
t
i
l
t
a
nd
s
l
a
nt
,
P
os
i
t
i
o
n
of
fa
c
e
i
n
i
m
a
g
e
,
Im
a
g
e
l
i
gh
t
i
ng
v
a
ri
a
t
i
on,
E
xpr
e
s
s
i
on
V
a
ri
a
t
i
on)
.
A
s
s
how
n
i
n
F
i
gur
e
5.
F
i
gu
r
e
5.
E
xa
m
p
l
e
s
o
f
f
a
c
e
i
m
a
ge
s
o
f
t
w
o
i
ndi
vi
dua
l
s
f
r
o
m
t
h
e
F
a
c
e
s
96
da
t
a
s
e
t
3
.
4
.
D
atab
as
e
d
e
s
c
r
i
p
ti
on
F
a
c
e
s
96
W
e
us
e
d
da
t
a
b
a
s
e
a
r
e
fe
r
e
n
c
e
F
a
c
e
s
96
i
n
[
18
]
w
h
i
c
h
c
on
t
a
i
n
s
152
s
ubj
e
c
t
s
a
nd
e
a
c
h
s
ub
j
e
c
t
c
on
t
a
i
ns
20
d
i
ffe
re
n
t
fa
c
e
s
of
t
h
e
s
a
m
e
p
e
rs
on.
T
he
r
e
s
ol
u
t
i
o
n
of
a
l
l
8
-
bi
t
i
m
a
ge
s
i
s
196x
196
(V
a
r
i
a
t
i
on
o
f
i
nd
i
vi
du
a
l
'
s
i
m
a
ge
s
B
a
c
kgroun
ds
).
H
e
a
d
S
c
a
l
e
,
H
e
a
d
t
urn
,
t
i
l
t
a
n
d
s
l
a
nt
,
P
os
i
t
i
o
n
of
f
a
c
e
i
n
i
m
a
g
e
,
Im
a
g
e
l
i
ght
i
ng
v
a
r
i
a
t
i
on
,
E
xpr
e
s
s
i
on
V
a
ri
a
t
i
on:
s
om
e
e
xpr
e
s
s
i
on
v
a
ri
a
t
i
on
,
A
d
di
t
i
ona
l
c
om
m
e
nt
,
i
m
a
ge
s
w
e
re
t
a
ke
n
i
n
a
s
i
n
gl
e
s
e
s
s
i
on.
A
s
s
how
n
i
n
F
i
gure
6.
F
i
gure
6.
E
xa
m
p
l
e
s
of
f
a
c
e
i
m
a
g
e
s
of
t
w
o
i
ndi
vi
du
a
l
s
fro
m
t
he
F
a
c
e
s
96
da
t
a
s
e
t
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
69
5
-
7
04
700
4.
A
LG
O
R
I
T
H
M
P
R
O
P
O
S
ED
T
hi
s
a
l
g
ori
t
hm
i
s
b
a
s
e
d
on
our
w
o
rk
i
n
[
19
,
20]
.
W
e
de
c
o
m
po
s
e
t
h
e
d
a
t
a
ba
s
e
i
nt
o
t
w
o
da
t
a
b
a
s
e
s
w
h
i
c
h
a
re
t
h
e
t
e
s
t
da
t
a
b
a
s
e
a
nd
t
h
e
t
ra
i
ni
ng
d
a
t
a
ba
s
e
by
di
ffe
r
e
nt
p
e
rc
e
nt
a
ge
s
.
F
or
e
x
a
m
pl
e
,
50%
t
e
s
t
b
a
s
e
a
nd
50%
t
ra
i
ni
ng
b
a
s
e
,
40
%
t
e
s
t
ba
s
e
a
nd
60%
t
r
a
i
n
i
ng
ba
s
e
a
nd
30
%
t
e
s
t
ba
s
e
a
nd
7
0%
t
r
a
i
n
i
ng
ba
s
e
.
T
h
e
n
be
t
w
e
e
n
t
he
t
w
o
i
m
a
g
e
s
,
on
e
b
e
l
on
gs
t
o
t
h
e
t
e
s
t
b
a
s
e
a
nd
t
h
e
ot
he
r
t
o
t
he
t
ra
i
ni
n
g
ba
s
e
,
t
he
n
e
xt
ra
c
t
t
he
r
e
m
a
rk
a
bl
e
po
i
nt
s
for
e
a
c
h
on
e
by
rob
us
t
m
e
t
hods
(S
IF
T
,
P
CA
-
S
IF
T
,
A
S
IF
T
,
a
nd
S
U
RF
)
a
ft
e
r
n
orm
a
l
i
z
i
ng
da
t
a
t
h
e
n
s
e
l
e
c
t
s
t
he
b
e
s
t
po
i
nt
s
by
t
h
e
RA
N
S
A
C
a
l
gori
t
h
m
.
T
he
n
d
e
t
e
r
m
i
n
e
s
t
h
e
s
i
m
i
l
a
r
po
i
nt
s
by
t
he
m
e
t
r
i
c
of
t
h
e
E
uc
l
i
d
e
a
n
di
s
t
a
n
c
e
be
t
w
e
e
n
t
w
o
v
e
c
t
ors
.
T
h
e
s
t
e
ps
of
t
he
pro
pos
e
d
m
e
t
h
od
s
u
m
m
a
ri
z
e
i
n
t
h
e
a
l
gor
i
t
h
m
be
l
ow
a
s
s
how
n
i
n
F
i
g
ure
7
.
T
h
e
propos
e
d
t
e
c
hni
q
ue
(
t
h
e
po
i
nt
n
orm
a
l
i
z
a
t
i
on
d
e
t
e
c
t
e
d
by
S
U
RF
a
nd
t
h
e
n
t
he
a
s
s
oc
i
a
t
i
on
w
i
t
h
t
h
e
RA
N
S
A
C
a
l
gor
i
t
hm
)
i
s
i
l
l
us
t
r
a
t
e
d
i
n
t
he
F
i
gure
8
b
e
l
ow
.
T
h
e
v
a
l
i
da
t
i
on
of
o
ur
e
xp
e
ri
e
nc
e
a
l
l
ow
s
us
t
o
ha
ve
t
e
xt
by
fou
r
d
a
t
a
ba
s
e
s
(O
R
L
,
F
a
c
e
s
95,
F
a
c
e
96
,
a
nd
G
r
i
m
a
c
e
)
a
c
c
ord
i
ng
t
o
d
i
ffe
re
n
t
va
r
i
a
t
i
ons
of
f
a
c
e
s
.
S
t
a
rt
.
S
t
e
p
1
:
In
p
u
t
t
w
o
i
m
a
g
e
s
o
n
e
b
e
l
o
n
g
i
n
g
t
o
t
h
e
t
e
s
t
d
a
t
a
b
a
s
e
a
n
d
t
h
e
o
t
h
e
r
t
o
t
h
e
t
ra
i
n
i
n
g
d
a
t
a
b
a
s
e
S
t
e
p
2
:
E
x
t
ra
c
t
t
h
e
re
m
a
rk
a
b
l
e
p
o
i
n
t
s
o
f
t
h
e
fa
c
e
b
y
ro
b
u
s
t
m
e
t
h
o
d
s
(S
IF
T
,
P
CA
-
S
IF
T
,
A
S
I
F
T
,
a
n
d
S
U
R
F
).
S
t
e
p
3
:
N
o
rm
a
l
i
z
e
d
a
t
a
d
e
t
e
c
t
s
S
t
e
p
4
:
Id
e
n
t
i
fi
c
a
t
i
o
n
o
f
o
u
t
l
i
e
rs
b
y
t
h
e
RA
N
S
A
C
a
l
g
o
ri
t
h
m
,
t
h
e
n
c
a
n
c
e
l
l
a
t
i
o
n
o
f
t
h
e
s
e
p
o
i
n
t
s
t
o
e
s
t
a
b
l
i
s
h
t
h
e
c
o
rre
c
t
m
a
t
c
h
i
n
g
S
t
e
p
5
:
Ca
l
c
u
l
a
t
e
t
h
e
re
c
o
g
n
i
t
i
o
n
ra
t
e
b
y
t
h
e
fo
rm
u
l
a
o
ft
e
n
.
100
.
c
orre
c
t
m
atc
he
s
R
e
c
ognition
rate
x
total
m
atc
he
s
=
(6
)
S
t
e
p
6
:
Re
p
e
a
t
t
h
e
s
t
e
p
s
fo
r
e
a
c
h
i
m
a
g
e
i
n
t
h
e
q
u
e
ry
s
e
t
a
n
d
g
a
l
l
e
ry
a
p
p
l
y
t
h
e
d
i
ffe
re
n
t
d
e
t
e
c
t
o
rs
t
o
a
s
s
i
g
n
a
s
i
m
u
l
a
t
e
d
i
m
a
g
e
S
t
o
p
.
F
i
g
ure
7
.
B
a
s
i
c
s
t
e
ps
i
nvo
l
ve
d
i
n
t
h
e
a
ppl
i
c
a
t
i
on
of
t
h
e
pr
opo
s
e
d
m
e
t
hod
Ide
nt
i
c
a
l
fa
c
e
F
i
gure
8
.
I
l
l
us
t
ra
t
i
on
of
t
h
e
propos
e
d
m
e
t
hod
b
y
d
a
t
a
ba
s
e
s
O
RL
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
Ne
w
appr
oa
c
h
t
o
t
h
e
i
de
nt
i
f
i
c
at
i
on
of
t
h
e
e
as
y
e
x
pr
e
s
s
i
on
r
e
c
ogni
t
i
on
s
y
s
t
e
m
…
(
A
hm
e
d
C
hat
e
r
)
701
4.
1
.
S
ome
s
i
mu
l
at
i
on
r
e
s
u
l
ts
an
d
d
i
s
c
u
s
s
i
on
In
t
h
i
s
s
e
c
t
i
on
,
w
e
h
a
ve
s
t
ud
i
e
d
t
he
di
f
fe
r
e
nt
robus
t
d
e
t
e
c
t
o
rs
us
i
ng
t
h
e
(S
IF
T
,
P
CA
-
S
IF
T
,
A
S
IF
T
)
[
21
,
22
,
23,
24]
a
nd
S
U
RF
m
e
t
h
ods
by
va
ry
i
ng
t
h
e
e
xpre
s
s
i
o
n
of
t
he
fa
c
e
s
of
t
h
e
s
a
m
e
pe
rs
on.
T
h
e
n
,
w
e
ha
v
e
m
e
a
s
ure
d
t
he
nu
m
be
r
of
de
s
c
ri
p
t
ors
,
t
he
nu
m
be
r
of
m
a
t
c
h
e
s
a
nd
t
h
e
proc
e
s
s
i
ng
t
i
m
e
b
y
d
i
ffe
re
n
t
d
e
t
e
c
t
ors
a
c
c
ordi
ng
t
o
t
h
e
c
h
a
ng
e
i
n
fa
c
i
a
l
e
xp
re
s
s
i
on
.
U
s
i
ng
t
h
e
da
t
a
b
a
s
e
i
m
a
g
e
s
of
r
e
a
l
f
a
c
e
s
,
t
he
F
i
gu
re
s
9
-
14
b
e
l
ow
s
how
s
om
e
e
xa
m
pl
e
s
of
t
h
e
e
x
t
ra
c
t
i
on
of
d
e
s
c
ri
pt
ors
a
nd
s
i
m
i
l
a
r
po
i
nt
s
by
d
i
ff
e
re
n
t
robus
t
d
e
t
e
c
t
ors
.
A
f
t
e
r
ha
vi
n
g
s
e
v
e
r
a
l
s
i
m
u
l
a
t
i
ons
t
e
xt
s
,
w
e
fi
nd
t
h
a
t
t
h
e
d
e
t
e
c
t
ors
(A
S
IF
T
)
e
xc
e
e
d
t
h
e
d
e
t
e
c
t
or
(S
IF
T
)
i
n
t
e
r
m
s
of
de
s
c
r
i
pt
or
.
A
s
i
l
l
us
t
ra
t
e
d
i
n
t
he
F
ig
ur
e
11.
F
i
g
ure
9
.
T
he
t
w
o
l
e
f
t
f
a
c
e
s
t
he
e
xt
r
a
c
t
i
on
by
fe
a
t
ure
A
S
IF
T
a
nd
t
h
e
t
w
o
r
i
gh
t
f
a
c
e
s
by
f
e
a
t
ure
S
IF
T
F
i
gure
10
.
F
e
a
t
ure
e
xt
r
a
c
t
i
on
by
P
CA
-
S
IF
T
a
nd
S
U
RF
F
i
gure
11
.
Corre
s
p
onde
nc
e
b
e
t
w
e
e
n
l
e
a
rni
n
g
a
nd
e
v
a
l
u
a
t
i
on
fa
c
e
s
,
i
n
t
h
e
fi
rs
t
l
i
ne
(S
IF
T
)
a
nd
t
h
e
s
e
c
o
nd
l
i
n
e
(A
S
IF
T
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
69
5
-
7
04
702
T
he
s
i
m
u
l
a
t
i
on
r
e
s
ul
t
s
how
s
t
h
a
t
t
he
A
S
IF
T
de
t
e
c
t
or
gi
v
e
s
t
h
e
nu
m
b
e
r
of
d
e
s
c
r
i
pt
ors
h
i
gh
e
r
t
h
a
n
S
IF
T
a
s
i
l
l
us
t
r
a
t
e
d
i
n
F
i
gure
11
.
H
e
re
a
ft
e
r,
w
e
us
e
t
he
P
CA
-
S
IF
T
de
t
e
c
t
or
,
i
n
F
ig
ure
12
w
h
i
c
h
a
l
l
ow
s
s
i
gni
f
i
c
a
nt
i
m
pro
ve
m
e
n
t
i
n
t
e
r
m
s
of
c
o
rre
s
pon
di
ng
p
ow
e
r
.
T
h
e
s
i
m
ul
a
t
i
on
r
e
s
ul
t
s
b
y
t
h
e
propos
e
d
m
e
t
h
od.
W
e
d
i
d
s
om
e
t
e
s
t
s
on
t
he
b
a
s
i
s
(O
RL
,
F
a
c
e
s
95
,
F
a
c
e
s
96
,
a
nd
G
ri
m
a
c
e
).
F
i
rs
t
t
e
s
t
:
t
he
t
w
o
fa
c
e
s
of
di
ffe
r
e
nt
p
e
op
l
e
,
t
he
F
ig
ure
13
,
b
e
l
ow
s
how
s
t
h
e
re
s
u
l
t
.
W
h
e
n
w
e
p
ut
t
w
o
d
i
ff
e
re
n
t
fa
c
e
s
,
w
e
do
no
t
fi
n
d
a
ny
s
i
m
i
l
a
r
de
s
c
r
i
pt
or,
a
s
i
l
l
us
t
r
a
t
e
d
i
n
t
he
F
i
g
u
r
e
1
3
.
S
e
c
on
d
t
e
s
t
:
t
h
e
t
w
o
f
a
c
e
s
of
t
he
s
a
m
e
pe
rs
on
be
l
ong
t
o
four
da
t
a
b
a
s
e
s
w
i
t
h
di
ffe
r
e
n
t
p
o
s
i
t
i
on
,
F
i
gure
14
b
e
l
ow
s
how
s
t
he
re
s
u
l
t
.
F
i
gure
12
.
P
CA
-
S
IF
T
m
a
t
c
he
s
be
t
w
e
e
n
l
e
a
rni
ng
a
nd
e
v
a
l
ua
t
i
on
f
a
c
e
s
F
i
gure
13
.
0
k
e
y
poi
n
t
m
a
t
c
he
s
nor
m
a
l
i
z
a
t
i
on
by
S
U
RF
+
RA
N
S
A
C
w
i
t
h
d
i
ffe
r
e
n
t
f
a
c
e
s
F
i
gure
14
.
M
a
t
c
he
s
of
ke
y
po
i
nt
s
nor
m
a
l
i
z
e
by
S
U
RF
+
RA
N
S
A
C
V
a
l
i
d
a
t
i
on
of
o
ur
a
pp
roa
c
h
ba
s
e
d
on
a
c
o
m
pa
r
i
s
on
b
e
t
w
e
e
n
t
h
e
fol
l
ow
i
ng
m
e
t
hods
(S
IF
T
,
P
CA
-
S
IF
T
,
A
S
IF
T
)
a
nd
t
he
pr
opos
e
d
t
e
c
hn
i
que
,
w
e
h
a
v
e
c
l
a
r
i
fi
e
d
our
m
e
t
hod
,
i
nc
r
e
a
s
e
d
t
h
e
re
c
ogn
i
t
i
on
r
a
t
e
c
o
m
pa
r
e
d
t
o
ot
he
r
e
x
i
s
t
e
d
t
e
c
h
ni
qu
e
s
.
T
h
e
re
s
u
l
t
s
o
f
our
s
i
m
u
l
a
t
i
on
c
a
n
be
s
um
m
a
ri
z
e
d
i
n
t
he
T
a
bl
e
1
b
y
four
da
t
a
b
a
s
e
s
.
T
he
r
e
s
ul
t
s
of
ou
r
da
t
a
b
a
s
e
(F
a
c
e
s
95)
s
i
m
u
l
a
t
i
on
c
a
n
b
e
s
um
m
a
r
i
z
e
d
i
n
t
he
T
a
b
l
e
2
.
T
h
e
re
s
ul
t
s
of
our
da
t
a
ba
s
e
(F
a
c
e
s
9
6)
s
i
m
u
l
a
t
i
on
c
a
n
be
s
um
m
a
r
i
z
e
d
i
n
t
h
e
T
a
b
l
e
3
.
T
he
re
s
ul
t
s
of
our
da
t
a
b
a
s
e
(G
ri
m
a
c
e
)
s
i
m
ul
a
t
i
on
c
a
n
be
s
u
m
m
a
ri
z
e
d
i
n
t
he
T
a
bl
e
4.
T
h
e
t
e
s
t
o
n
t
he
four
da
t
a
ba
s
e
s
,
a
s
i
l
l
us
t
r
a
t
e
d
i
n
t
he
t
a
bl
e
s
a
bov
e
,
s
how
s
t
ha
t
o
ur
m
e
t
hod
off
e
rs
s
a
t
i
s
f
a
c
t
ory
re
s
u
l
t
s
i
n
t
e
rm
s
of
re
c
ogn
i
t
i
on
ra
t
e
s
a
s
s
how
n
i
n
T
a
b
l
e
5
.
T
a
b
l
e
1.
T
he
s
i
m
ul
a
t
i
on
re
s
u
l
t
i
n
t
e
rm
s
of
t
h
e
a
v
e
ra
g
e
of
t
he
de
t
e
c
t
e
d
de
s
c
ri
pt
ors
,
t
he
a
ve
r
a
ge
of
c
orre
c
t
m
a
t
c
h
e
s
,
re
c
ogn
i
t
i
on
a
c
c
ur
a
c
y
on
d
a
t
a
ba
s
e
(O
R
L
)
M
e
t
h
o
d
T
h
e
a
v
e
ra
g
e
o
f
t
h
e
d
e
t
e
c
t
e
d
d
e
s
c
ri
p
t
o
rs
T
h
e
a
v
e
ra
g
e
o
f
c
o
rre
c
t
m
a
t
c
h
e
s
Re
c
o
g
n
i
t
i
o
n
a
c
c
u
ra
c
y
(%
)
S
IF
T
40
3
7
.
6
94
A
S
IF
T
60
57
95
P
CA
-
S
IF
T
20
1
9
.
1
2
9
5
.
6
T
h
e
m
e
t
h
o
d
p
ro
p
o
s
e
s
30
29
96.
6
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
Ne
w
appr
oa
c
h
t
o
t
h
e
i
de
nt
i
f
i
c
at
i
on
of
t
h
e
e
as
y
e
x
pr
e
s
s
i
on
r
e
c
ogni
t
i
on
s
y
s
t
e
m
…
(
A
hm
e
d
C
hat
e
r
)
703
T
a
b
l
e
2
.
T
he
s
i
m
ul
a
t
i
on
re
s
u
l
t
i
n
t
e
rm
s
of
t
h
e
a
v
e
ra
g
e
of
t
he
de
t
e
c
t
e
d
de
s
c
ri
pt
ors
,
t
he
a
ve
r
a
ge
of
c
orre
c
t
m
a
t
c
h
e
s
,
re
c
ogn
i
t
i
on
a
c
c
ur
a
c
y
on
d
a
t
a
ba
s
e
(
F
a
c
e
s
95)
M
e
t
h
o
d
T
h
e
a
v
e
ra
g
e
o
f
t
h
e
d
e
t
e
c
t
e
d
d
e
s
c
ri
p
t
o
rs
T
h
e
a
v
e
ra
g
e
o
f
c
o
rre
c
t
m
a
t
c
h
e
s
Re
c
o
g
n
i
t
i
o
n
a
c
c
u
ra
c
y
(%
)
S
IF
T
45
3
7
.
6
9
5
.
1
6
A
S
IF
T
70
67
9
5
.
7
3
P
CA
-
S
IF
T
25
2
4
.
1
3
9
5
.
5
2
T
h
e
m
e
t
h
o
d
p
ro
p
o
s
e
s
36
3
4
.
7
7
96.
6
T
a
b
l
e
3
.
T
he
s
i
m
ul
a
t
i
on
re
s
u
l
t
i
n
t
e
rm
s
of
t
h
e
a
v
e
ra
g
e
of
t
he
de
t
e
c
t
e
d
de
s
c
ri
pt
ors
,
t
he
a
ve
r
a
ge
of
c
orre
c
t
m
a
t
c
h
e
s
,
re
c
ogn
i
t
i
on
a
c
c
ur
a
c
y
on
d
a
t
a
ba
s
e
(
F
a
c
e
s
96)
M
e
t
h
o
d
T
h
e
a
v
e
ra
g
e
o
f
t
h
e
d
e
t
e
c
t
e
d
d
e
s
c
ri
p
t
o
rs
T
h
e
a
v
e
ra
g
e
o
f
c
o
rre
c
t
m
a
t
c
h
e
s
Re
c
o
g
n
i
t
i
o
n
a
c
c
u
ra
c
y
(%
)
S
IF
T
48
4
5
.
9
5
9
5
.
7
4
A
S
IF
T
80
7
6
.
7
2
9
5
,
9
P
CA
-
S
IF
T
37
3
5
.
3
7
9
5
.
6
T
h
e
m
e
t
h
o
d
p
ro
p
o
s
e
s
28
2
7
.
1
6
97
T
a
b
l
e
4
.
T
he
s
i
m
ul
a
t
i
on
re
s
u
l
t
i
n
t
e
rm
s
of
t
h
e
a
v
e
ra
g
e
of
t
he
de
t
e
c
t
e
d
de
s
c
ri
pt
ors
,
t
he
a
ve
r
a
ge
of
c
orre
c
t
m
a
t
c
h
e
s
,
re
c
ogn
i
t
i
on
a
c
c
ur
a
c
y
on
d
a
t
a
ba
s
e
(G
r
i
m
a
c
e
)
M
e
t
h
o
d
T
h
e
a
v
e
ra
g
e
o
f
t
h
e
d
e
t
e
c
t
e
d
d
e
s
c
ri
p
t
o
rs
T
h
e
a
v
e
ra
g
e
o
f
c
o
rre
c
t
m
a
t
c
h
e
s
Re
c
o
g
n
i
t
i
o
n
a
c
c
u
ra
c
y
(%
)
S
IF
T
36
3
4
.
1
8
94
.
9
7
A
S
IF
T
74
7
1
.
4
8
9
6
,
6
P
CA
-
S
IF
T
18
17.
22
9
5
.
6
7
T
h
e
m
e
t
h
o
d
p
ro
p
o
s
e
s
25
24
97
T
a
b
l
e
5
.
A
v
e
ra
g
e
pro
c
e
s
s
i
ng
t
i
m
e
be
t
w
e
e
n
t
w
o
i
m
a
g
e
s
of
t
h
e
s
a
m
e
fa
c
e
a
c
c
ordi
n
g
t
o
di
ff
e
re
nt
va
r
i
a
n
t
s
,
a
ppl
i
c
a
t
i
on
on
t
h
e
fo
l
l
ow
i
ng
da
t
a
b
a
s
e
s
(O
RL
,
F
a
c
e
s
96
,
F
a
c
e
s
96
,
a
nd
G
RIM
A
C
E
)
M
e
t
h
o
d
S
IF
T
P
CA
-
S
IF
T
A
S
IF
T
T
h
e
m
e
t
h
o
d
p
ro
p
o
s
e
s
A
v
e
ra
g
e
t
i
m
e
o
f
c
o
m
p
a
ri
s
o
n
b
e
t
w
e
e
n
t
w
o
i
m
a
g
e
s
(s
)
b
y
(O
RL
)
0
.
7
8
0
.
6
9
1
.
9
5
6
0
.
3
6
3
A
v
e
ra
g
e
t
i
m
e
o
f
c
o
m
p
a
r
i
s
o
n
b
e
t
w
e
e
n
t
w
o
i
m
a
g
e
s
(s
)
b
y
F
a
c
e
s
9
6
0
.
7
8
0
.
7
4
1
.
5
8
6
0
.
5
7
2
A
v
e
ra
g
e
t
i
m
e
o
f
c
o
m
p
a
ri
s
o
n
b
e
t
w
e
e
n
t
w
o
i
m
a
g
e
s
(s
)
b
y
G
ri
m
a
c
e
1
.
2
0
.
9
1
.
8
0
.
6
7
5
A
v
e
ra
g
e
t
i
m
e
o
f
c
o
m
p
a
ri
s
o
n
b
e
t
w
e
e
n
t
w
o
i
m
a
g
e
s
(s
)
b
y
F
a
c
e
s
9
6
0
.
5
4
0
.
5
2
1
.
4
0
6
0
.
4
9
3
A
ft
e
r
c
om
p
a
ri
ng
our
m
e
t
hod
w
i
t
h
ot
h
e
r
t
e
c
hn
i
qu
e
s
,
w
e
f
i
nd
t
ha
t
o
ur
m
e
t
ho
d
out
p
e
rfor
m
s
t
h
e
ot
h
e
rs
i
n
t
e
r
m
s
of
t
i
m
e
t
o
i
d
e
nt
i
fy
t
h
e
c
orr
e
s
pond
i
ng
f
a
c
e
,
a
s
s
how
n
i
n
t
h
e
T
a
bl
e
5
.
T
he
r
e
s
ul
t
s
of
ou
r
s
i
m
u
l
a
t
i
on
s
how
t
ha
t
t
h
e
prop
os
e
d
m
e
t
hod
r
e
s
ponds
f
a
s
t
e
r
t
h
a
n
ot
he
r
de
t
e
c
t
ors
,
a
nd
gi
v
e
goo
d
re
s
u
l
t
s
by
d
i
ff
e
re
n
t
c
ha
ng
e
s
,
a
s
i
t
i
s
s
how
n
i
n
t
h
e
pr
e
vi
ous
s
i
m
u
l
a
t
i
on
.
S
o
t
h
a
t
w
e
c
a
n
a
pp
l
y
t
o
t
r
a
c
k
m
ovi
n
g
ob
j
e
c
t
s
i
n
re
a
l
t
i
m
e
.
I
n
t
he
fu
t
ur
e
w
orks
fol
l
ow
w
e
t
e
xt
our
m
e
t
h
od
on
t
he
fo
l
l
ow
i
ng
da
t
a
b
a
s
e
publ
i
c
.
Ck
[25
]
,
O
ul
u
-
CA
S
IA
[2
6].
4
C
O
N
C
LU
S
I
O
N
In
t
h
i
s
a
rt
i
c
l
e
,
w
e
h
a
ve
c
om
p
a
r
e
d
t
he
rob
us
t
d
e
t
e
c
t
ors
w
i
t
h
t
he
prop
os
e
d
m
e
t
ho
d
t
o
d
e
t
e
r
m
i
n
e
t
he
d
i
ff
e
re
n
t
p
a
ra
m
e
t
e
rs
.
T
he
t
e
s
t
i
s
m
a
d
e
on
t
h
e
four
da
t
a
ba
s
e
s
(O
RL
,
F
a
c
e
s
95
,
F
a
c
e
96
,
a
nd
G
ri
m
a
c
e
)
.
T
he
s
i
m
u
l
a
t
i
on
r
e
s
ul
t
s
s
how
t
ha
t
t
h
e
propos
e
d
m
e
t
ho
d
(t
h
e
po
i
nt
no
rm
a
l
i
z
a
t
i
on
de
t
e
c
t
s
by
S
U
RF
a
n
d
t
h
e
n
t
he
a
s
s
oc
i
a
t
i
on
w
i
t
h
t
h
e
RA
N
S
A
C
a
l
gori
t
hm
a
l
gor
i
t
h
m
)
g
i
ve
s
good
re
s
u
l
t
s
i
n
t
e
r
m
s
of
de
t
e
r
m
i
n
i
ng
s
i
m
i
l
a
r
de
s
c
r
i
pt
ors
by
d
i
ff
e
re
n
t
v
a
ri
a
t
i
ons
i
n
f
a
c
i
a
l
pos
e
,
re
c
ogn
i
t
i
on
r
a
t
e
a
n
d
a
ve
r
a
ge
t
i
m
e
of
c
o
m
p
a
ri
s
on
be
t
w
e
e
n
t
w
o,
one
b
e
l
o
ngs
t
o
t
he
t
e
s
t
d
a
t
a
ba
s
e
a
n
d
t
he
ot
h
e
r
o
ne
t
o
t
he
t
ra
i
ni
ng
d
a
t
a
ba
s
e
,
c
o
m
p
a
re
d
t
o
(S
IF
T
,
A
S
IF
T
,
PAC
-
S
IF
T
).
T
h
i
s
a
ppr
oa
c
h
t
ha
t
w
e
propos
e
d
g
a
ve
good
re
s
ul
t
s
i
n
t
e
r
m
s
of
re
c
ogn
i
t
i
on
r
a
t
e
a
n
d
i
m
a
g
e
a
na
l
ys
i
s
t
i
m
e
.
S
o
,
t
hi
s
a
pp
roa
c
h
c
a
n
a
na
l
yz
e
i
m
a
g
e
s
i
n
r
e
a
l
t
i
m
e
.
In
t
h
e
fut
ur
e
w
ork
,
t
h
e
e
v
a
l
u
a
t
i
on
of
o
ur
t
e
c
hn
i
qu
e
i
s
a
ppl
i
c
a
t
e
d
a
t
a
s
e
t
a
t
pub
l
i
c
.
C
k
, O
u
l
u
-
CA
S
IA
c
o
m
p
a
re
d
t
o
t
he
re
s
u
l
t
found
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
1693
-
6930
T
E
L
K
O
M
N
IK
A
T
e
l
e
c
om
m
un
Co
m
put
E
l
Con
t
rol
,
V
ol
.
18
,
N
o.
2
,
A
pri
l
2
020:
69
5
-
7
04
704
R
EF
ER
EN
C
ES
[
1]
H
-
F
.
H
ua
n
g
a
nd
S
-
C
.
T
a
i
,
“
F
a
c
i
a
l
e
xp
r
e
s
s
i
on
r
e
c
og
ni
t
i
on
u
s
i
ng
ne
w
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
on
a
l
g
or
i
t
hm
,
”
E
l
e
c
t
r
on.
L
e
t
t
.
C
om
put
.
V
i
s
.
I
m
age
A
na
l
.
,
vo
l
.
1
1,
no
.
1
,
p
p.
41
–
54
,
2012
.
[
2]
Y
a
ng,
M
i
n
g
-
H
s
ua
n
.
"
K
e
r
ne
l
E
i
ge
nf
a
c
e
s
v
s
.
K
e
r
ne
l
F
i
s
h
e
r
f
a
c
e
s
:
F
a
c
e
R
e
c
ogn
i
t
i
on
U
s
i
ng
K
e
r
n
e
l
M
e
t
hods
.
"
F
gr
.
V
ol
.
2
.
200
2.
[
3]
M
.
A
.
T
ur
k
a
nd
A
.
P
.
P
e
n
t
l
a
nd
,
“
F
a
c
e
r
e
c
o
gni
t
i
on
u
s
i
n
g
e
i
ge
n
f
a
c
e
s
,
”
I
n
P
r
oc
.
I
E
E
E
C
on
f
.
C
om
pu
t
e
r
V
i
s
i
on
a
nd
P
at
t
e
r
n
R
e
c
og
ni
t
i
on
,
p
p.
5
8
6
-
591
,
1991
.
[
4]
J
.
Y
a
ng
,
D
.
Z
h
a
ng
,
A
.
F
.
F
r
a
ng
i
,
a
nd
J
.
Y
a
ng
,
“
T
w
o
-
di
m
e
ns
i
on
a
l
P
C
A
:
A
ne
w
a
ppr
oa
c
h
t
o
a
pp
e
a
r
a
n
c
e
-
ba
s
e
d
f
a
c
e
r
e
p
r
e
s
e
n
t
a
t
i
o
n
a
nd
r
e
c
o
gn
i
t
i
on
,
”
I
E
E
E
T
r
a
ns
a
c
t
i
on
s
on
P
a
t
t
e
r
n
A
n
al
y
s
i
s
and
M
ac
h
i
n
e
I
n
t
e
l
l
i
ge
nc
e
,
vo
l
.
26,
n
o.
1
,
pp.
13
1
-
1
37,
2
004
.
[
5]
D
.
G
.
L
ow
e
,
“
D
i
s
t
i
nc
e
i
m
a
g
e
f
e
a
t
u
r
e
s
f
r
o
m
s
c
a
l
e
-
i
n
va
r
i
a
nt
ke
ypo
i
nt
s
,
”
I
nt
.
J
ou
r
na
l
of
C
om
p
ut
e
r
V
i
s
i
o
n
,
vo
l
.
60,
no.
2
,
pp.
9
1
-
1
10,
2
004
.
[
6]
J
.
L
uo
,
Y
.
M
a
,
E
.
T
a
ki
k
a
w
a
,
S
.
H
.
L
a
o
,
M
.
K
a
w
a
de
,
a
nd
B
.
L
.
L
u,
“
P
e
r
s
on
-
s
p
e
c
i
f
i
c
S
I
F
T
f
e
a
t
ur
e
s
f
or
f
a
c
e
r
e
c
o
gn
i
t
i
on
,
”
I
n
t
e
r
na
t
i
ona
l
C
on
f
e
r
e
n
c
e
on
A
c
ou
s
t
i
c
,
S
pe
e
c
h
a
nd
Si
gnal
P
r
o
c
e
s
s
i
ng
(
I
C
A
SSP
2
007)
,
H
a
w
a
i
i
,
v
ol
.
2
,
pp.
59
3
-
5
96,
2
007
.
[
7]
X
.
Q
u
e
t
a
l
.
,
“
E
va
l
u
a
t
i
on
of
S
I
F
T
a
nd
S
U
R
F
f
or
vi
s
i
on
b
a
s
e
d
l
oc
a
l
i
z
a
t
i
on,
”
I
n
t
e
r
nat
i
o
nal
A
r
c
hi
v
e
s
o
f
t
he
P
ho
t
o
gr
am
m
e
t
r
y
,
R
e
m
ot
e
S
e
ns
i
n
g
a
nd
Spa
t
i
al
I
nf
or
m
a
t
i
on
Sc
i
e
nc
e
s
,
vo
l
.
41
-
B
3
,
pp
.
685
-
692
,
201
6.
[
8]
P
.
M
.
K
u
m
a
r
e
t
al
.
,
“
I
n
t
e
l
l
i
g
e
nt
f
a
c
e
r
e
c
o
gni
t
i
on
a
nd
na
vi
ga
t
i
o
n
s
y
s
t
e
m
us
i
ng
ne
ur
a
l
l
e
a
r
ni
n
g
f
or
s
m
a
r
t
s
e
c
u
r
i
t
y
i
n
I
nt
e
r
n
e
t
o
f
T
hi
ngs
,
”
C
l
us
t
e
r
C
om
put
i
ng
,
vo
l
.
2
2,
no
.
4,
p
p.
77
33
-
77
44,
20
17.
[
9]
W
an
g,
H
ao
,
e
t
a
l
.
"C
o
s
f
ac
e
:
L
ar
ge
m
ar
g
i
n
c
o
s
i
ne
l
o
s
s
f
or
de
e
p
f
ac
e
r
e
c
o
gni
t
i
on
.
"
P
r
o
c
e
e
di
ngs
o
f
t
he
I
E
E
E
C
onf
e
r
e
n
c
e
on
C
om
p
ut
e
r
V
i
s
i
on
and
P
a
t
t
e
r
n
R
e
c
og
ni
t
i
on.
2
018
.
[
10]
A
.
C
ha
t
e
r
a
n
d
A
.
L
a
s
f
a
r
,
“
D
e
t
e
c
t
i
on
o
f
i
m
a
g
e
de
s
c
r
i
p
t
o
r
s
a
nd
m
od
i
f
i
c
a
t
i
on
of
t
h
e
w
e
i
gh
t
i
ng
f
unc
t
i
on
f
or
t
h
e
e
s
t
i
m
a
t
i
o
n
of
t
h
e
f
unda
m
e
nt
a
l
m
a
t
r
i
x
us
i
n
g
r
o
bus
t
m
e
t
ho
ds
,
”
J
ou
r
na
l
o
f
E
n
gi
n
e
e
r
i
n
g
and
A
pp
l
i
e
d
S
c
i
e
nc
e
s
,
vol
.
13,
n
o.
7
,
pp.
1
835
-
184
3,
20
18
.
[
11]
M
.
G
h
e
r
ghe
r
e
h
c
hi
,
S
.
Y
.
K
i
m
,
H
.
A
f
a
r
i
de
h
,
a
nd
Y
.
S
.
K
i
m
,
“
R
A
N
d
om
s
a
m
pl
e
c
o
ns
e
ns
u
s
(
R
A
N
S
A
C
)
a
l
go
r
i
t
h
m
f
o
r
e
nha
n
c
i
ng
ove
r
l
a
pp
e
d
e
t
c
h
e
d
t
r
a
c
k
c
oun
t
i
ng
,
”
I
E
T
I
m
age
P
r
oc
e
s
s
.
,
vol
.
9,
no
.
2,
p
p.
97
–
106
,
201
5.
[
12]
H
.
B
a
y
,
A
.
E
s
s
,
T
.
T
u
yt
e
l
a
a
r
s
,
L
.
V
a
n
G
ool
,
"
S
p
e
e
d
e
d
-
up
r
obu
s
t
f
e
a
t
ur
e
s
(
S
U
R
F
)
,
"
C
o
m
p
ut
.
V
i
s
.
I
m
a
ge
U
nde
r
s
t
.
,
110(
3)
,
346
-
359
(
2008
)
.
[
13]
L
.
S
h
a
o
e
t
a
l
.
,
“
S
pa
t
i
o
-
t
e
m
p
or
a
l
L
a
p
l
a
c
i
a
n
py
r
a
m
i
d
c
o
di
n
g
f
o
r
a
c
t
i
on
r
e
c
ogn
i
t
i
on
,
”
I
E
E
E
T
r
a
ns
a
c
t
i
o
ns
o
n
C
ybe
r
ne
t
i
c
s
,
v
ol
.
44
,
no.
6
,
pp
.
817
-
827
,
201
4.
[
14]
A
.
C
ha
t
e
r
a
nd
A
.
L
a
s
f
a
r
,
“
R
obu
s
t
H
a
r
r
i
s
de
t
e
c
t
o
r
c
o
r
r
e
s
p
ond
i
ng
a
n
d
c
a
l
c
ul
a
t
e
s
t
h
e
p
r
o
j
e
c
t
i
on
e
r
r
or
us
i
ng
t
he
m
od
i
f
i
c
a
t
i
on
of
t
h
e
w
e
i
g
ht
i
ng
f
u
nc
t
i
o
n,
”
I
nt
e
r
n
at
i
on
al
J
o
ur
na
l
of
M
a
c
hi
ne
L
e
a
r
ni
ng
and
C
om
pu
t
i
ng
(
I
J
M
L
C
)
,
vol
.
9,
no
.
1,
p
p.
62
-
66
,
201
9.
[
15]
A
T
&
T
D
a
t
a
ba
s
e
of
F
a
c
e
s
‘
O
R
L
F
a
c
e
D
a
t
a
ba
s
e
’
A
T
&
T
L
a
bo
r
a
t
o
r
i
e
s
,
C
a
m
b
r
i
dge
:
ht
t
p
:
/
/
c
a
m
-
or
l
.
c
o.
u
k/
f
a
c
e
d
a
t
a
ba
s
e
.
ht
m
l
[
16]
L
i
bo
r
s
pa
c
e
k
’
s
F
a
c
i
a
l
i
m
a
ge
da
t
a
b
a
s
e
s
‘
G
r
i
m
a
c
e
f
a
c
e
D
a
t
a
b
a
s
e
:
ht
t
p
:
/
/
c
s
w
w
w
.
e
s
s
e
x
.
a
c
.
u
k/
m
v
/
a
l
l
f
a
c
e
s
/
G
r
i
m
a
c
e
.
h
t
m
l
[
17]
L
i
bo
r
L
i
bo
r
s
pa
c
e
k
’
s
F
a
c
i
a
l
i
m
a
ge
da
t
a
b
a
s
e
s
‘
f
a
c
e
95
I
m
a
ge
D
a
t
a
ba
s
e
:
ht
t
p
:
/
/
c
s
w
w
w
.
e
s
s
e
x
.
a
c
.
uk
/
m
v
/
a
l
l
f
a
c
e
s
/
a
c
e
9
5.
h
t
m
l
[
18]
L
i
bo
r
L
i
bor
s
pa
c
e
k
’
s
F
a
c
i
a
l
i
m
a
g
e
da
t
a
ba
s
e
s
‘
f
a
c
e
9
6
I
m
a
ge
D
a
t
a
b
a
s
e
:
h
t
t
p:
/
/
c
s
w
w
w
.
e
s
s
e
x
.
a
c
.
uk
/
m
v/
a
l
l
f
a
c
e
s
/
a
c
e
9
6.
h
t
m
l
[
19]
A
.
C
ha
t
e
r
a
nd
A
.
L
a
s
f
a
r
,
“
C
om
pa
r
i
s
on
o
f
r
ob
us
t
m
e
t
h
ods
f
or
e
x
t
r
a
c
t
i
ng
de
s
c
r
i
p
t
o
r
s
a
nd
f
a
c
i
a
l
m
a
t
c
hi
n
g,
”
I
nt
e
r
n
at
i
on
al
C
on
f
e
r
e
nc
e
o
n
W
i
r
e
l
e
s
s
T
e
c
hno
l
og
i
e
s
,
E
m
be
dde
d
and
I
nt
e
l
l
i
ge
n
t
Sy
s
t
e
m
s
(
W
I
T
S)
,
M
or
o
c
c
o
,
pp.
1
-
4,
2
019
.
[
20]
A
.
C
ha
t
e
r
a
nd
A
.
L
a
s
f
a
r
,
“
N
e
w
a
ppr
oa
c
h
t
o
c
a
l
c
ul
a
t
i
ng
t
h
e
f
und
a
m
e
nt
a
l
m
a
t
r
i
x
,
”
I
n
t
.
J
.
E
l
e
c
t
r
.
C
om
p
ut
.
E
ng.
I
J
E
C
E
,
vol
.
10,
n
o.
3
,
pp.
2
357
–
2366
,
J
u
n.
20
20.
[
21]
A
.
V
i
na
y,
V
.
S
.
S
he
kha
r
,
A
.
K
um
a
r
C
.
,
S
.
N
a
t
a
r
a
j
a
n,
a
nd
K
.
N
.
B
.
M
ur
t
h
y,
“
A
f
f
i
ne
-
s
c
a
l
e
i
n
va
r
i
a
nt
f
e
a
t
u
r
e
t
r
a
n
s
f
or
m
a
nd
t
w
o
-
d
i
m
e
n
s
i
ona
l
pr
i
n
c
i
pa
l
c
o
m
po
ne
n
t
a
n
a
l
ys
i
s
:
A
nove
l
f
r
a
m
e
w
or
k
f
o
r
a
f
f
i
ne
a
nd
s
c
a
l
e
i
nva
r
i
a
n
t
f
a
c
e
r
e
c
o
gn
i
t
i
on
,
”
I
E
T
C
om
put
e
r
V
i
s
i
on
,
vol
.
10
,
no
.
1
,
pp
.
43
-
59
,
2
016
.
[
22]
D
.
M
i
s
hk
i
n
,
J
.
M
a
t
a
s
,
a
nd
M
.
P
e
r
d
oc
h
,
“
M
o
ds
:
F
a
s
t
a
nd
r
ob
us
t
m
e
t
hod
f
or
t
w
o
-
v
i
e
w
m
a
t
c
hi
ng,
”
C
om
put
e
r
V
i
s
i
on
and
I
m
ag
e
U
n
de
r
s
t
and
i
ng
,
vo
l
.
14
1,
pp
.
81
-
93
,
201
5.
[
23]
J
i
a
ng,
D
a
Y
ou
,
a
n
d
J
on
gw
e
o
n
K
i
m
.
"
A
r
t
w
o
r
k
p
a
i
nt
i
ng
i
de
nt
i
f
i
c
a
t
i
o
n
m
e
t
hod
f
or
pa
n
or
a
m
a
ba
s
e
d
o
n
a
d
a
pt
i
v
e
r
e
c
t
i
l
i
ne
a
r
pr
oj
e
c
t
i
o
n
a
nd
op
t
i
m
i
z
e
d
A
S
I
F
T
.
"
M
u
l
t
i
m
e
d
i
a
T
o
ol
s
a
nd
A
pp
l
i
c
a
t
i
o
ns
78
.
22
,
pp
.
3189
3
-
3
1924
,
201
9
.
[
24]
S
a
bha
r
w
a
l
,
T
a
nu
pr
e
e
t
,
e
t
a
l
.
"
R
e
c
ogn
i
t
i
o
n
of
s
u
r
g
i
c
a
l
l
y
a
l
t
e
r
e
d
f
a
c
e
i
m
a
ge
s
:
a
n
e
m
p
i
r
i
c
a
l
a
n
a
l
y
s
i
s
on
r
e
c
e
n
t
a
dva
n
c
e
s
.
"
A
r
t
i
f
i
c
i
a
l
I
nt
e
l
l
i
g
e
nc
e
R
e
v
i
e
w
52
.
2
,
p
p.
10
09
-
1040
,
201
9
.
[
25]
T
.
K
a
n
a
de
,
J
.
F
.
C
ohn
,
a
nd
Y
.
T
i
a
n,
“
C
o
m
pr
e
he
ns
i
ve
d
a
t
a
ba
s
e
f
or
f
a
c
i
a
l
e
x
pr
e
s
s
i
on
a
na
l
y
s
i
s
,
”
P
r
o
c
e
e
di
n
gs
F
o
ur
t
h
I
E
E
E
I
n
t
e
r
na
t
i
o
na
l
C
o
nf
e
r
e
nc
e
on
A
u
t
om
a
t
i
c
F
ac
e
and
G
e
s
t
u
r
e
R
e
c
ogn
i
t
i
on
,
pp
.
46
-
53
,
2
000
.
[
26]
G
.
Z
h
a
o,
X
.
H
ua
ng,
M
.
T
a
i
ni
,
S
.
Z
.
L
i
,
a
nd
M
.
P
i
e
t
i
k
a
i
ne
n
,
“
F
a
c
i
a
l
e
x
pr
e
s
s
i
on
r
e
c
ogn
i
t
i
o
n
f
r
o
m
ne
a
r
-
i
n
f
r
a
r
e
d
v
i
d
e
os
,
”
I
m
age
an
d
v
i
s
i
o
n
C
om
pu
t
i
n
g
,
vo
l
.
29,
n
o.
9
,
pp.
6
07
-
619
,
2001
.
Evaluation Warning : The document was created with Spire.PDF for Python.