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
.
4
,
A
u
g
u
s
t
201
9
,
p
p
.
2
4
5
1
~2
4
5
9
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v9
i
4
.
p
p
2
4
5
1
-
2459
2451
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
Cro
w
d
recog
nition sy
ste
m
bas
ed o
n optica
l f
lo
w
along
w
ith
SVM
clas
sifier
Sh
re
eda
rsh
a
n
K
.
1
,
S.
Set
h
u Selv
i
2
De
p
a
rtme
n
t
o
f
El
e
c
tro
n
ics
a
n
d
C
o
m
m
u
n
ica
ti
o
n
E
n
g
in
e
e
rin
g
,
M
S
Ra
m
a
iah
In
sti
tu
te
o
f
T
e
c
h
n
o
l
o
g
y
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
11
,
2
0
1
7
R
ev
i
s
ed
Dec
2
8
,
2
0
1
7
A
cc
ep
ted
Mar
4
,
2
0
1
9
T
h
e
m
a
n
u
sc
rip
t
d
isc
u
ss
e
s
a
b
o
u
t
a
b
n
o
rm
a
li
ti
e
s
in
a
c
ro
w
d
e
d
sc
e
n
a
rio
.
T
o
p
re
v
e
n
t
th
e
m
ish
a
p
a
t
a
p
u
b
li
c
p
la
c
e
,
th
e
re
is
n
o
m
u
c
h
m
e
c
h
a
n
is
m
wh
ich
c
o
u
l
d
p
re
v
e
n
t
o
r
a
lert
th
e
c
o
n
c
e
rn
e
d
a
u
t
h
o
rit
y
a
b
o
u
t
su
s
p
e
c
ts
in
a
c
ro
w
d
.
Us
u
a
ll
y
in
a
c
ro
w
d
e
d
sc
e
n
e
,
th
e
re
a
r
e
c
h
a
n
c
e
s
o
f
so
m
e
m
ish
a
p
li
k
e
a
terro
rist
a
tt
a
c
k
o
r
a
c
rime
.
Ou
r
targ
e
t
is
f
in
d
in
g
tec
h
n
i
q
u
e
s
t
o
id
e
n
t
ify
su
c
h
a
c
ti
v
it
ies
a
n
d
t
o
p
o
ss
ib
ly
p
re
v
e
n
t
t
h
e
m
.
If
th
e
c
ro
w
d
m
e
m
b
e
rs
e
x
h
ib
it
a
b
n
o
rm
a
l
b
e
h
a
v
io
r,
w
e
c
o
u
ld
i
d
e
n
ti
f
y
a
n
d
sa
y
th
a
t
th
is
p
a
rti
c
u
lar
p
e
rso
n
is
a
su
sp
e
c
t
a
n
d
t
h
e
n
th
e
c
o
n
c
e
rn
e
d
a
u
th
o
rit
y
w
o
u
ld
l
o
o
k
i
n
to
th
e
m
a
tt
e
r.
T
h
e
re
a
re
v
a
rio
u
s
m
e
th
o
d
s to
id
e
n
ti
f
y
th
e
a
b
n
o
rm
a
l
b
e
h
a
v
io
r.
T
h
e
p
ro
p
o
se
d
a
p
p
ro
a
c
h
is
b
a
se
d
o
n
o
p
ti
c
a
l
f
lo
w
m
o
d
e
l.
It
h
a
s
a
n
a
b
il
it
y
to
d
e
tec
t
th
e
su
d
d
e
n
c
h
a
n
g
e
s
in
m
o
ti
o
n
o
f
a
n
in
d
iv
id
u
a
l
a
m
o
n
g
th
e
c
ro
w
d
.
F
irs
t,
th
e
m
a
in
re
g
io
n
o
f
m
o
ti
o
n
is
e
x
trac
ted
b
y
th
e
h
e
l
p
o
f
m
o
ti
o
n
h
e
a
t
m
a
p
.
Ha
rris
c
o
rn
e
r
d
e
tec
to
r
is
u
se
d
f
o
r
e
x
trac
ti
n
g
p
o
in
t
o
f
in
tere
st
o
f
e
x
trac
ted
m
o
ti
o
n
a
r
e
a
.
Ba
se
d
o
n
th
e
p
o
in
t
o
f
i
n
tere
st
a
n
o
p
ti
c
a
l
f
lo
w
is
e
sti
m
a
ted
h
e
re
.
Af
ter
a
n
a
ly
z
in
g
th
is
o
p
t
ica
l
f
lo
w
m
o
d
e
l,
a
th
re
sh
o
ld
v
a
lu
e
is
f
ix
e
d
.
Ba
sic
a
ll
y
o
p
ti
c
a
l
f
lo
w
is
a
n
e
n
e
rg
y
lev
e
l
o
f
in
d
iv
id
u
a
l
f
ra
m
e
.
T
h
e
th
re
sh
o
l
d
v
a
lu
e
is
f
o
rw
a
rd
e
d
to
S
V
M
c
las
sif
ier,
w
h
ich
p
ro
d
u
c
e
s
a
b
e
tt
e
r
re
su
lt
w
it
h
9
9
.
7
1
%
a
c
c
u
ra
c
y
.
T
h
i
s
a
p
p
r
o
a
c
h
is
v
e
ry
u
se
f
u
l
in
re
a
l
t
im
e
v
id
e
o
su
rv
e
il
lan
c
e
s
y
ste
m
w
h
e
re
a
m
a
c
h
in
e
c
a
n
m
o
n
it
o
r
u
n
w
a
n
ted
c
ro
w
d
a
c
ti
v
it
y
.
K
ey
w
o
r
d
s
:
A
b
n
o
r
m
al
b
eh
a
v
io
u
r
Mo
tio
n
h
ea
t
m
ap
Op
tical
f
lo
w
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
:
Sh
r
ee
d
ar
s
h
a
n
K
.
,
Dep
ar
t
m
en
t o
f
E
lectr
o
n
ics a
n
d
C
o
m
m
u
n
icat
io
n
E
n
g
i
n
ee
r
in
g
,
M
S R
a
m
aiah
I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
,
MSR
I
T
P
o
s
t,
B
an
g
alo
r
e
-
5
6
0
0
5
4
,
Kar
n
atak
a,
I
n
d
ia
.
E
m
ail:
s
h
r
ee
d
ar
s
h
a
n
.
k
@
m
s
r
i
t.e
d
u
1.
I
NT
RO
D
UCT
I
O
N
Den
s
i
t
y
o
f
p
o
p
u
latio
n
i
n
u
r
b
a
n
ar
ea
s
is
m
u
ch
h
i
g
h
er
co
m
p
ar
ed
to
r
u
r
al
ar
ea
s
.
So
,
th
er
e
i
s
a
h
i
g
h
er
o
cc
u
r
r
en
ce
o
f
ab
n
o
r
m
al
ac
ti
v
itie
s
s
u
ch
a
s
ille
g
al
g
at
h
er
i
n
g
,
g
r
o
u
p
f
ig
h
ti
n
g
etc.
T
h
es
e
ac
tiv
itie
s
m
a
y
b
e
d
an
g
er
o
u
s
f
o
r
o
u
r
s
o
cial
p
u
b
l
ic
s
ec
u
r
it
y
.
T
o
d
etec
t
s
u
ch
ill
eg
al
ac
ti
v
ities
,
cr
o
w
d
b
e
h
av
i
o
r
p
r
e
d
ictio
n
is
an
in
ter
esti
n
g
to
p
ic
o
f
a
v
id
eo
m
o
n
ito
r
in
g
s
y
s
te
m
an
d
h
as
a
g
r
e
at
v
alu
e
f
o
r
s
t
u
d
y
an
d
r
esear
c
h
.
I
t
is
v
er
y
d
i
f
f
ic
u
lt
to
an
al
y
ze
in
d
i
v
id
u
al
cr
o
w
d
m
e
m
b
er
s
w
h
et
h
er
th
eir
b
eh
av
io
r
is
n
o
r
m
al
o
r
ab
n
o
r
m
al.
T
h
er
e
ar
e
s
o
m
e
f
ac
to
r
s
lik
e
n
u
m
b
er
o
f
cr
o
w
d
o
b
j
ec
ts
,
m
o
tio
n
o
f
cr
o
w
d
o
b
j
ec
ts
,
an
d
s
eg
r
eg
atio
n
o
f
cr
o
w
d
o
b
j
ec
ts
,
eli
m
i
n
ati
n
g
t
h
e
b
ac
k
g
r
o
u
n
d
d
is
t
u
r
b
an
ce
,
ca
tc
h
th
e
v
ar
ian
ce
a
m
o
n
g
cr
o
w
d
o
b
j
ec
t e
tc.
,
w
h
ich
is
r
eq
u
ir
ed
to
an
al
y
ze
an
d
p
r
ed
ict
th
e
o
u
tco
m
e.
So
m
e
s
c
h
o
lar
s
h
av
e
m
ad
e
s
ev
er
al
ac
h
iev
e
m
e
n
ts
i
n
th
i
s
ar
ea
.
T
o
d
escr
ib
e
th
e
cr
o
w
d
b
eh
av
io
r
W
an
g
et
al
.
[
1
]
h
as
i
m
p
le
m
en
ted
an
i
m
p
r
o
v
ed
s
p
atial
-
te
m
p
o
r
al
f
ea
tu
r
e
w
it
h
ad
ap
tiv
e
s
iz
e.
A
g
r
ad
ien
t
b
ased
s
p
atial
m
o
d
el
is
u
s
ed
to
d
ef
in
e
th
e
m
o
tio
n
o
f
o
b
j
ec
ts
in
s
ce
n
es
an
d
r
ec
o
g
n
ize
th
e
t
y
p
e
o
f
lo
ca
l
cr
o
w
d
b
eh
av
io
r
b
y
Kr
atz
et
al
.
[
2
]
.
L
B
P
-
T
O
P
alg
o
r
ith
m
i
s
i
m
p
le
m
e
n
ted
b
y
X
u
et
al
.
[
3
]
to
ex
tr
ac
t
th
e
cr
o
w
d
b
eh
av
io
r
.
T
h
is
alg
o
r
ith
m
m
ak
e
s
a
L
D
A
m
o
d
el
b
y
th
e
h
elp
o
f
tr
ain
i
n
g
an
d
b
ased
o
n
th
at
m
o
d
el
it d
ec
id
es th
e
lo
ca
l c
r
o
w
d
b
eh
av
io
r
.
T
h
e
ab
o
v
e
d
is
cu
s
s
ed
m
o
d
el
m
ai
n
l
y
i
m
p
le
m
e
n
ts
s
p
a
tial
-
te
m
p
o
r
al
f
ea
tu
r
e,
g
r
ad
ien
t f
ea
t
u
r
e
o
r
tr
ain
in
g
tex
t
u
r
e
f
ea
t
u
r
e
to
d
eter
m
in
e
th
e
lo
ca
l
cr
o
w
d
b
eh
a
v
io
r
.
I
n
all
th
ese
m
o
d
els
t
h
er
e
is
a
lack
o
f
p
r
o
v
id
in
g
in
f
o
r
m
atio
n
o
f
g
lo
b
al
c
h
ar
ac
ter
is
tics
o
f
t
h
e
cr
o
w
d
,
a
n
d
h
en
ce
th
e
co
m
p
lete
i
n
f
o
r
m
atio
n
o
f
t
h
e
cr
o
w
d
ca
n
n
o
t
b
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
87
08
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t 2
0
1
9
:
2
4
5
1
-
2459
2452
o
b
tain
ed
.
So
it
is
i
m
p
o
r
ta
n
t
to
u
n
d
er
s
ta
n
d
th
e
cr
o
w
d
b
eh
av
io
r
al
o
f
co
m
p
lete
s
ce
n
e
w
i
th
o
u
t
k
n
o
w
in
g
t
h
e
in
d
iv
id
u
al
ac
t
io
n
.
R
ec
en
t
l
y
s
o
m
e
m
o
d
els
ar
e
p
r
o
p
o
s
ed
w
h
ic
h
a
r
e
b
ased
o
n
Hid
d
en
m
ar
k
o
v
m
o
d
el
(
HM
M)
[
4
,
5
]
,
L
ag
r
an
g
ia
n
C
o
h
er
en
t
Stru
ct
u
r
es
(
L
C
S)
[
6
]
,
So
cial
Fo
r
ce
Mo
d
el
(
SF
M)
[
7
]
,
MA
P
R
M
[
8
]
,
C
h
ao
tic
I
n
v
ar
ia
n
ts
(
C
I
)
[
9
]
an
d
Kin
e
m
atic
Featu
r
e
(
K
F)
[
1
0
]
.
T
h
ese
m
o
d
el
s
ca
n
g
i
v
e
b
etter
p
er
f
o
r
m
an
ce
in
d
etec
tio
n
o
f
cr
o
w
d
b
eh
av
io
r
al
o
n
s
o
m
e
g
i
v
en
b
en
ch
m
ar
k
[
7
,
9
]
.
B
u
t th
es
e
m
eth
o
d
s
m
a
y
f
ail
to
d
etec
t
th
e
cr
o
w
d
b
eh
av
io
r
al
in
ca
s
e
o
f
lo
w
r
eso
l
u
tio
n
v
id
eo
o
r
if
th
e
s
p
ee
d
o
f
o
b
j
ec
t
is
to
o
f
ast
o
r
t
o
o
s
lo
w
.
I
n
[
1
1
]
,
a
m
et
h
o
d
is
p
r
o
p
o
s
ed
to
o
b
tain
b
eh
a
v
io
u
r
al
r
ec
o
g
n
itio
n
wh
ich
is
b
a
s
ed
o
n
Vio
le
n
ce
Fl
o
w
s
th
a
t
s
tar
t
s
f
r
o
m
co
m
p
u
tatio
n
o
f
o
p
tical
f
lo
w
b
et
w
ee
n
ad
j
ac
en
t
f
r
a
m
e
s
.
T
h
is
m
et
h
o
d
is
w
el
l
s
u
itab
le
f
o
r
a
ch
o
s
en
v
id
eo
s
et
b
u
t
it n
ee
d
s
m
o
r
e
en
h
a
n
ce
m
en
t
s
t
o
g
et
m
o
r
e
ac
cu
r
ate
r
esu
lt.
I
n
th
is
p
ap
er
,
w
e
h
a
v
e
d
em
o
n
s
tr
ated
an
ap
p
r
o
ac
h
th
at
is
b
ased
o
n
o
p
tical
f
lo
w
p
atter
n
.
So
m
e
tr
ea
t
m
en
t
h
a
s
b
ee
n
d
o
n
e
alo
n
g
w
it
h
o
p
tical
f
lo
w
p
atter
n
in
o
r
d
e
r
to
d
etec
t
th
e
cr
o
w
d
b
eh
av
io
r
.
T
o
ex
tr
ac
t th
e
r
eg
io
n
o
f
m
o
tio
n
ac
tiv
it
y
,
w
e
ta
k
e
th
e
h
elp
o
f
m
o
tio
n
h
ea
t
m
ap
i
n
a
cr
o
w
d
s
c
en
e.
T
h
is
h
ea
t
m
ap
r
ed
u
ce
s
th
e
ti
m
e
o
f
p
r
o
ce
s
s
i
n
g
an
d
g
iv
e
s
b
etter
r
esu
lt
t
h
at
p
l
a
y
s
a
v
er
y
i
m
p
o
r
tan
t
r
o
le
in
r
ea
l
ti
m
e
ap
p
licatio
n
.
A
h
o
t
r
e
g
io
n
o
f
t
h
e
s
ce
n
e
g
iv
es
t
h
e
p
o
in
t
o
f
i
n
ter
est
ea
s
i
l
y
.
T
h
ese
p
o
in
ts
o
f
in
ter
e
s
t
m
ak
e
a
b
o
u
n
d
ar
y
in
h
o
t
ar
ea
s
o
f
th
e
s
ce
n
e
an
d
b
ased
o
n
th
e
m
ar
k
ed
ar
ea
,
o
p
tical
f
lo
w
is
co
m
p
u
ted
.
Op
tical
f
lo
w
i
n
f
o
r
m
atio
n
m
ak
e
s
a
m
u
lti
m
o
d
el
o
p
tical
f
lo
w
p
atte
r
n
in
d
i
f
f
er
e
n
t ti
m
e
b
ased
o
n
t
h
e
b
eh
a
v
io
r
al
o
f
th
e
cr
o
w
d
.
V
ar
iatio
n
s
o
f
m
o
tio
n
s
ar
e
esti
m
ated
to
m
ak
e
a
d
is
tin
ctio
n
in
f
a
v
o
r
o
f
p
o
ten
tial
ab
n
o
r
m
al
ev
e
n
t
s
.
W
e
c
o
n
s
id
er
th
at
w
e
ar
e
in
a
c
r
o
w
d
en
v
ir
o
n
m
e
n
t
w
i
th
n
o
r
estrictio
n
s
o
f
th
e
n
u
m
b
er
o
f
p
eo
p
le.
T
o
an
al
y
ze
t
h
e
o
p
tical
f
lo
w
m
o
d
el,
w
e
h
av
e
ta
k
e
n
s
tan
d
ar
d
UM
N
d
ataset
an
d
p
er
f
o
r
m
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
T
h
e
p
ap
er
is
o
r
g
an
ized
is
as
f
o
l
lo
w
:
I
n
s
ec
tio
n
2
a
b
r
ief
r
ela
ted
w
o
r
k
is
p
r
esen
ted
.
Sec
tio
n
3
ta
lk
s
ab
o
u
t
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
.
W
e
co
m
p
ar
e
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
w
ith
Sp
ar
s
e
R
ec
o
n
s
tr
u
ctio
n
C
o
s
t
(
SR
C
)
[
1
]
,
C
h
ao
tic
I
n
v
ar
ia
n
ts
(
C
I
)
[
1
2
]
,
So
cial
Fo
r
ce
Mo
d
el
(
SF
)
[
4
]
an
d
th
e
f
o
r
ce
f
ie
ld
m
o
d
el
(
FF
)
[
1
3
]
o
n
UM
N
d
ataset.
I
n
s
ec
tio
n
4
o
u
tco
m
e
is
e
v
alu
a
ted
an
d
an
al
y
s
is
i
s
p
er
f
o
r
m
ed
.
L
ast
s
ec
tio
n
co
n
clu
d
es t
h
e
p
ap
er
.
2.
RE
L
AT
E
D
WO
RK
L
i
u
et
al
.
[
1
4
]
h
as
p
r
o
j
ec
ted
an
ap
p
r
o
ac
h
i.e
.
ag
en
t
-
b
ased
m
o
ti
o
n
m
o
d
els
(
A
MM
s
)
w
h
ic
h
u
s
e
s
m
u
ltip
le
ex
e
m
p
lar
to
d
etec
t
cr
o
w
d
b
e
h
a
v
io
r
o
f
ca
p
tu
r
ed
s
ce
n
e
tr
aj
ec
to
r
y
.
T
h
e
y
p
r
o
p
o
s
ed
an
o
p
tim
iza
tio
n
al
g
o
r
ith
m
t
h
at
co
r
r
elate
s
b
et
w
ee
n
tr
aj
ec
to
r
y
d
ata
an
d
ex
e
m
p
lar
A
MM
t
h
at
is
b
ased
o
n
KL
-
d
i
v
er
g
e
n
ce
an
d
E
x
ten
d
ed
Kal
m
a
n
S
m
o
o
th
er
.
I
n
o
r
d
er
to
d
escr
ib
e
th
e
cr
o
w
d
m
o
tio
n
,
th
e
y
i
n
tr
o
d
u
ce
d
th
e
n
o
v
el
in
d
i
v
id
u
al
f
ea
tu
r
e
al
o
n
g
w
it
h
h
o
lis
tic
f
ea
t
u
r
e
th
at
i
s
b
ased
o
n
p
r
o
p
o
s
ed
m
ea
s
u
r
e
m
e
n
t
o
f
co
r
r
elatio
n
.
T
h
e
r
esu
lt
p
r
o
v
es
th
at
t
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
is
w
ell
s
u
itab
le
f
o
r
r
ec
o
g
n
izi
n
g
r
ea
l
w
o
r
ld
cr
o
w
d
e
d
s
ce
n
es.
I
n
[
1
5
]
,
B
er
a
et
al
.
h
as p
r
o
p
o
s
ed
an
alg
o
r
it
h
m
f
o
r
lo
w
to
m
e
d
iu
m
d
e
n
s
i
t
y
cr
o
w
d
t
h
at
u
s
e
s
a
tr
aj
ec
to
r
y
b
eh
av
io
r
lear
n
i
n
g
p
r
o
ce
s
s
.
T
h
i
s
alg
o
r
it
h
m
co
m
b
i
n
es
tr
ac
k
i
n
g
alg
o
r
ith
m
,
m
o
d
el
o
f
n
o
n
-
li
n
ea
r
p
ed
estrian
m
o
t
io
n
an
d
tech
n
iq
u
es
o
f
B
a
y
esia
n
lear
n
in
g
to
co
m
p
u
te
t
h
e
b
eh
av
i
o
u
r
o
f
ea
ch
p
ed
estrian
in
d
iv
id
u
all
y
i
n
th
e
s
ce
n
e
.
T
h
ese
co
m
b
i
n
atio
n
s
o
f
f
ea
t
u
r
es
ar
e
u
s
ed
to
id
en
ti
f
y
t
h
e
d
i
f
f
er
en
t
m
o
tio
n
p
atter
n
s
o
f
p
ed
estrian
an
d
s
e
g
m
en
t
t
h
e
tr
aj
ec
to
r
ies.
P
E
T
S
2
0
1
6
A
R
E
NA
d
ata
s
et
h
a
s
b
ee
n
u
s
ed
f
o
r
t
h
is
a
lg
o
r
it
h
m
.
T
h
i
s
alg
o
r
it
h
m
i
s
s
u
itab
le
f
o
r
in
d
o
o
r
o
r
o
u
td
o
o
r
cr
o
w
d
s
ce
n
e
an
d
h
as g
o
t a
b
etter
p
er
f
o
r
m
a
n
ce
.
I
n
[
1
6
]
au
th
o
r
h
as p
r
o
p
o
s
ed
a
n
o
v
el
alg
o
r
it
h
m
to
d
etec
t a
b
n
o
r
m
al
cr
o
w
d
b
eh
av
io
u
r
w
h
ic
h
is
b
ased
o
n
th
e
ac
ce
ler
atio
n
f
ea
t
u
r
e.
T
h
is
alg
o
r
ith
m
e
x
p
lo
r
es
th
e
g
lo
b
al
f
ea
tu
r
e
o
f
th
e
c
u
r
r
en
t
b
e
h
av
i
o
u
r
al
s
tate
an
d
t
h
e
p
r
ev
io
u
s
b
eh
a
v
io
r
al
s
tate
u
n
l
i
k
e
p
r
ev
io
u
s
w
o
r
k
th
at
e
x
p
lo
r
es
o
n
l
y
lo
ca
l
f
ea
t
u
r
es.
I
n
t
h
i
s
p
ap
er
,
au
th
o
r
h
as
p
r
o
j
ec
ted
a
n
e
w
g
lo
b
al
ac
ce
ler
atio
n
f
ea
t
u
r
e
t
h
at
is
b
ased
o
n
in
v
ar
ia
n
ce
o
f
th
r
ee
co
n
s
ec
u
ti
v
e
f
r
a
m
es
o
f
g
r
e
y
-
s
ca
le
i
m
a
g
e
s
.
I
t
h
as
a
n
ab
ilit
y
t
o
m
a
tch
p
i
x
els
a
n
d
p
r
o
v
id
es
th
e
s
p
ee
d
ch
an
g
es
p
r
ec
is
el
y
.
Af
t
er
th
at,
an
lo
g
ar
it
h
m
o
f
d
etec
tio
n
i
s
d
esi
g
n
ed
w
it
h
t
h
e
h
elp
o
f
ac
ce
ler
atio
n
co
m
p
u
tatio
n
alo
n
g
w
ith
a
f
o
r
eg
r
o
u
n
d
s
tep
o
f
e
x
tr
ac
tio
n
.
T
h
is
alg
o
r
ith
m
i
s
r
o
b
u
s
t
b
ec
au
s
e
it
is
in
d
ep
en
d
e
n
t
o
f
d
etec
tio
n
an
d
s
eg
m
e
n
tatio
n
.
T
h
is
alg
o
r
ith
m
g
i
v
es
t
h
e
r
esu
lt
b
ased
o
n
th
e
t
h
r
esh
o
ld
an
al
y
s
is
.
Se
v
er
al
d
atasets
h
a
v
e
b
ee
n
tak
e
n
an
d
b
ased
o
n
th
at
r
esu
l
t
h
as
b
ee
n
co
m
p
u
ted
.
R
es
u
lt
s
o
b
tain
ed
ar
e
p
r
o
m
is
i
n
g
.
C
r
o
w
d
b
eh
av
io
r
al
d
etec
tio
n
is
an
ac
tiv
el
y
r
esear
ch
ed
ar
ea
w
h
er
e
b
eh
av
io
r
al
f
ea
t
u
r
es
ar
e
u
s
ed
to
d
if
f
er
e
n
tiate
b
et
w
ee
n
n
o
r
m
al
a
n
d
ab
n
o
r
m
al
ac
ti
v
ities
.
I
n
[
1
7
]
,
v
ar
io
u
s
t
y
p
e
s
o
f
f
ea
tu
r
es
ar
e
s
u
m
m
ar
ized
in
o
r
d
er
to
ca
p
tu
r
e
t
h
e
b
eh
a
v
io
r
o
f
th
e
c
r
o
w
d
.
T
h
ese
f
ea
tu
r
es
ar
e
p
r
o
j
e
cted
f
o
r
en
co
d
in
g
th
e
cr
o
w
d
b
e
h
av
io
r
,
w
h
ic
h
m
a
y
b
e
lo
ca
l
o
r
g
lo
b
al,
eith
er
tem
p
o
r
al
o
r
s
p
atial
o
r
b
o
th
.
P
o
s
itio
n
,
m
o
tio
n
b
y
o
p
tical
f
lo
w
[
1
8
]
,
f
o
r
eg
r
o
u
n
d
o
b
j
ec
t
s
ize
[
1
9
,
2
0
]
g
r
ad
ien
t
[
2
1
]
,
h
i
s
to
g
r
a
m
o
f
(
d
ir
ec
tio
n
,
m
o
tio
n
,
tex
t
u
r
e,
f
ield
)
[
2
2
]
,
tex
tu
r
e
a
n
d
te
m
p
o
r
al
s
p
atial
f
ea
t
u
r
es
ar
e
th
e
d
if
f
er
en
t
t
y
p
e
s
o
f
f
ea
t
u
r
es.
A
late
n
t
Dir
ic
h
let
a
llo
c
atio
n
m
o
d
el
is
in
tr
o
d
u
ce
d
i
n
[
2
3
]
f
o
r
c
r
o
w
d
ed
s
ce
n
e
w
h
er
e
m
o
tio
n
f
ea
t
u
r
e
o
f
d
ir
ec
tio
n
,
v
elo
cit
y
an
d
p
o
s
itio
n
is
u
s
ed
.
A
Ma
r
k
o
v
r
an
d
o
m
f
ield
m
o
d
el
is
p
r
o
p
o
s
ed
b
y
th
e
au
t
h
o
r
i
n
[
1
2
]
f
o
r
ab
n
o
r
m
a
l d
etec
ti
n
g
ac
ti
v
it
ies.
A
t e
ac
h
f
r
a
m
e
le
v
el,
o
p
tic
al
f
lo
w
f
ea
t
u
r
e
w
a
s
ex
tr
ac
ted
.
T
h
e
d
etec
tio
n
is
b
as
ed
o
n
th
e
p
r
o
b
ab
ilit
y
m
atr
i
x
.
T
h
ese
m
o
d
els
ar
e
w
e
ll
s
u
itab
l
e
f
o
r
s
ev
er
al
s
ce
n
es.
A
tr
aj
ec
to
r
y
is
e
x
tr
ac
ted
f
o
r
n
o
r
m
al
e
v
en
t
v
id
eo
s
an
d
i
s
co
m
p
ar
ed
w
it
h
ab
n
o
r
m
al
v
id
eo
s
in
[
2
4
]
,
s
o
th
at
th
e
ab
n
o
r
m
al
ac
ti
v
itie
s
ca
n
b
e
d
i
f
f
er
en
tia
ted
f
r
o
m
n
o
r
m
a
l
ac
ti
v
i
ties
.
I
n
[
2
5
]
au
t
h
o
r
h
as
g
e
n
er
ated
a
m
o
d
el
th
at
is
b
ased
o
n
m
i
x
tu
r
e
o
f
o
u
tlier
s
an
d
d
y
n
a
m
ic
te
x
t
u
r
es,
w
h
ich
is
lab
elled
as
an
o
m
a
lies
.
I
n
[
2
6
]
,
a
co
n
d
itio
n
al
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
C
r
o
w
d
r
ec
o
g
n
itio
n
s
ystem
b
a
s
ed
o
n
o
p
tica
l flo
w
a
lo
n
g
w
ith
S
V
M c
la
s
s
ifier
(
S
h
r
ee
d
a
r
s
h
a
n
.
K
)
2453
f
u
n
ctio
n
w
as
co
n
s
tr
u
cted
b
as
ed
o
n
th
e
f
ea
tu
r
e
o
f
d
ir
ec
tio
n
,
m
ag
n
it
u
d
e
an
d
p
o
s
itio
n
as
w
ell
as
a
B
a
y
e
s
ian
f
r
a
m
e
w
o
r
k
w
as
p
r
o
j
ec
ted
f
o
r
escap
in
g
d
etec
tio
n
t
h
r
o
u
g
h
m
o
d
ellin
g
cr
o
w
d
m
o
tio
n
i
n
b
o
th
t
h
e
ab
s
e
n
ce
an
d
p
r
esen
ce
o
f
escap
e
ev
e
n
t
s
.
Se
v
er
al
Sp
atio
-
te
m
p
o
r
al
t
y
p
es
ar
e
also
i
m
p
le
m
en
ted
.
I
n
o
r
d
er
to
id
en
tify
an
o
m
al
y
d
etec
tio
n
,
a
Sp
atio
-
te
m
p
o
r
al
t
y
p
e
b
lo
ck
w
i
s
e
ap
p
r
o
ac
h
is
u
s
ed
ap
p
ly
in
g
K
-
n
ea
r
e
s
t
n
ei
g
h
b
o
r
s
th
at
i
m
p
le
m
e
n
ts
co
-
o
cc
u
r
r
en
ce
m
o
d
el
[
13
-
27
]
.
A
cr
o
w
d
m
o
tio
n
p
atter
n
i
s
p
r
o
p
o
s
ed
in
[
2
8
]
f
o
r
cr
o
w
d
b
eh
a
v
i
o
r
al
d
etec
tio
n
w
h
ich
u
s
e
s
Sp
atio
-
te
m
p
o
r
al
t
y
p
es.
I
n
[
2
9
]
,
au
th
o
r
h
as
co
n
s
id
er
ed
b
o
th
te
m
p
o
r
al
an
d
s
p
atial
f
r
a
m
e
w
o
r
k
to
p
r
o
p
o
s
e
a
f
ea
t
u
r
e
th
at
is
b
ased
o
n
n
e
w
r
eg
io
n
in
o
r
d
er
to
d
escr
ib
e
ap
p
ea
r
an
ce
an
d
m
o
tio
n
o
f
b
o
th
th
e
f
r
a
m
e
w
o
r
k
.
B
o
th
s
p
atial
an
d
te
m
p
o
r
al
f
r
a
m
e
w
o
r
k
f
ea
t
u
r
e
is
d
eter
m
i
n
ed
to
g
en
er
ate
t
h
e
lo
ca
l
m
o
t
io
n
p
atter
n
b
e
h
a
v
io
r
[
2
2
]
ev
en
in
e
x
tr
e
m
el
y
cr
o
w
d
ed
s
c
en
ar
io
.
HM
M
f
ea
t
u
r
e
i
s
u
s
ed
t
o
d
eter
m
i
n
e
lo
ca
l
m
o
tio
n
p
att
er
n
th
at
is
b
ased
o
n
th
e
3
D
Gau
s
s
ian
d
is
tr
ib
u
tio
n
.
A
Sp
atio
-
te
m
p
o
r
al
m
o
d
el
is
also
i
m
p
le
m
e
n
ted
in
[
3
0
]
,
w
h
er
e
ab
n
o
r
m
al
ac
ti
v
itie
s
in
t
h
e
s
p
ee
d
,
s
ize
a
n
d
d
ir
ec
t
io
n
o
f
o
b
j
ec
t
w
er
e
d
etec
ted
.
A
v
id
eo
s
u
r
v
ei
llan
ce
s
y
s
te
m
is
d
e
v
elo
p
ed
in
[
3
1
-
3
2
]
f
o
r
ab
n
o
r
m
al
v
i
s
u
a
l d
etec
tio
n
an
d
r
ec
o
g
n
itio
n
in
cr
o
w
d
.
3.
P
RO
P
O
SE
D
SYS
T
E
M
3
.
1
.
M
o
t
io
n hea
t
m
a
p
T
h
is
w
o
r
k
Mo
tio
n
h
ea
t
m
ap
is
a
t
w
o
d
i
m
e
n
s
io
n
al
(
2
D)
g
r
ap
h
ical
r
ep
r
esen
tatio
n
o
f
an
i
m
ag
e
t
h
at
r
ep
r
esen
ts
a
h
i
s
to
g
r
a
m
w
h
er
e
m
o
s
t
ac
tiv
it
y
o
f
m
o
tio
n
o
cc
u
r
s
.
L
e
t
H
a
n
d
I
i
n
d
icate
t
h
e
h
ea
t
m
ap
a
n
d
i
n
te
n
s
it
y
o
f
f
r
a
m
e
r
esp
ec
tiv
el
y
.
(
,
)
=
−
1
(
,
)
+
(
,
)
(
1
)
0
=
0
(
,
)
(
2
)
W
h
er
e
n
an
d
(
n
-
1)
r
ep
r
esen
ts
t
h
e
cu
r
r
en
t a
n
d
th
e
p
r
ev
io
u
s
f
r
a
m
e
n
u
m
b
er
s
,
i a
n
d
j
ar
e
th
e
co
o
r
d
in
ates (
lin
e
a
n
d
co
lu
m
n
)
o
f
t
h
e
p
ix
el
(
i,j
)
o
f
a
f
r
a
m
e.
Hea
t
m
ap
is
j
u
s
t
u
s
ed
f
o
r
an
al
y
s
i
n
g
th
e
ar
ea
w
h
er
e
th
e
m
o
tio
n
ac
tiv
it
y
o
cc
u
r
s
.
T
h
is
w
i
ll
h
elp
in
r
ed
u
cin
g
p
r
o
ce
s
s
in
g
ti
m
e
a
n
d
i
m
p
r
o
v
es t
h
e
q
u
alit
y
o
f
t
h
e
r
es
u
lt.
3
.
2
.
P
o
ints o
f
inte
re
s
t
ex
t
ra
c
t
io
n
On
ce
h
ea
t
m
ap
is
g
e
n
er
ated
,
n
e
x
t
w
e
g
o
f
o
r
ex
tr
ac
ti
n
g
t
h
e
p
o
in
ts
o
f
in
ter
e
s
t.
Mo
r
av
ec
’
s
co
r
n
er
d
etec
to
r
is
a
s
i
m
p
le
alg
o
r
it
h
m
t
h
at
i
s
co
m
m
o
n
l
y
u
s
ed
f
o
r
th
is
p
u
r
p
o
s
e.
B
u
t
n
o
w
it
is
o
u
td
ated
.
So
m
et
i
m
es
it
w
a
s
n
o
t
ab
le
d
etec
t th
e
ac
cu
r
ate
co
r
n
er
ed
g
e
d
u
e
to
w
ea
k
i
n
v
ar
ian
t r
esp
o
n
s
e
w
it
h
r
esp
ec
t to
d
ir
ec
tio
n
.
S
o
it is
co
n
s
id
er
ed
as
n
o
is
e
s
e
n
s
iti
v
e.
Mo
r
av
ec
’
s
co
r
n
er
m
o
d
el
h
ad
m
i
n
i
m
al
co
m
p
lex
i
t
y
i
n
co
m
p
u
tatio
n
al
p
r
o
ce
s
s
,
b
u
t
th
er
e
ar
e
li
m
i
ta
tio
n
s
th
at
w
er
e
d
is
c
u
s
s
e
d
ea
r
lier
.
T
h
e
o
th
er
w
a
y
o
f
ex
tr
ac
tin
g
p
o
in
t
o
f
in
ter
est
is
Har
r
is
co
r
n
er
d
etec
to
r
m
o
d
el.
I
t
is
co
m
p
u
tatio
n
a
ll
y
d
em
a
n
d
i
n
g
a
n
d
o
v
er
co
m
e
s
all
th
e
li
m
itatio
n
s
o
f
Mo
r
av
ec
’
s
co
r
n
er
.
T
h
is
m
o
d
el
is
i
m
p
le
m
en
ted
in
t
h
e
c
u
r
r
en
t
p
ap
er
.
I
t
is
in
v
ar
ia
n
t
to
d
ir
ec
tio
n
an
d
ill
u
m
in
a
tio
n
v
ar
iatio
n
.
I
t
is
en
tire
l
y
b
ased
o
n
a
lo
ca
l
au
t
o
co
r
r
elate
d
s
ig
n
al
f
u
n
ctio
n
.
T
h
e
im
p
le
m
en
ted
f
u
n
ctio
n
ca
n
m
ea
s
u
r
e
th
e
s
i
g
n
al
w
it
h
h
elp
o
f
p
atch
s
h
if
t
in
v
ar
io
u
s
d
ir
ec
tio
n
s
,
th
e
s
h
i
f
t
b
ein
g
v
er
y
s
m
all.
L
et
u
s
ass
u
m
e
a
p
o
in
t
(
x
,
y
)
an
d
a
s
h
i
f
t
(
∆x
,
∆
y
)
t
h
e
n
th
e
lo
ca
l
au
to
-
co
r
r
elatio
n
f
u
n
ctio
n
is
d
ef
i
n
ed
as:
(
,
)
=
∑
[
(
,
)
−
(
+
∆
,
+
∆
)
]
2
(
3
)
w
h
er
e
I
d
en
o
tes
t
h
e
i
m
a
g
e
f
u
n
ctio
n
an
d
(
x
_
i,
y
_
i)
ar
e
th
e
p
o
i
n
ts
i
n
t
h
e
s
m
o
o
t
h
cir
cu
lar
w
i
n
d
o
w
w
ca
n
ter
ed
o
n
(
x
,
y
)
.
T
h
e
s
h
i
f
ted
i
m
a
g
e
is
ap
p
r
o
x
i
m
ated
b
y
a
T
ay
lo
r
ex
p
a
n
s
i
o
n
tr
u
n
ca
ted
to
th
e
f
ir
s
t o
r
d
er
t
er
m
s
as
(
+
∆
,
+
∆
)
≈
(
,
)
+
[
(
,
)
(
,
)
]
[
∆
∆
]
(
4
)
w
h
er
e
I
_
x
(
.
,
.
)
an
d
I
_
y
(
.
,
.
)
d
en
o
tes
th
e
p
ar
tial
d
er
iv
ati
v
e
s
in
x
an
d
y
,
r
esp
ec
tiv
el
y
.
S
u
b
s
ti
tu
tin
g
th
e
r
i
g
h
t
h
an
d
s
ite
o
f
(
4
)
in
to
(
3
)
y
ield
s
:
(
,
)
=
∑
(
[
(
,
)
(
,
)
]
[
∆
∆
]
)
2
(
5
)
=
[
∆
∆
]
(
,
)
[
∆
∆
]
(
6
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
87
08
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t 2
0
1
9
:
2
4
5
1
-
2459
2454
(
,
)
=
(
∑
(
(
,
)
)
2
∑
(
,
)
(
,
)
∑
(
,
)
(
,
)
∑
(
(
,
)
)
2
)
(
7
)
No
w
w
e
a
s
s
u
m
e
t
h
at
λ
1
an
d
λ
2
ar
e
th
e
E
ig
e
n
v
al
u
es
o
f
m
atr
ix
M
(
x
,
y
)
T
h
is
E
ig
e
n
v
alu
e
w
ill
d
escr
ib
e
o
u
r
p
o
in
t
o
f
in
ter
est.
I
f
th
e
v
alu
e
o
f
λ
1
an
d
λ
2
is
s
m
all,
th
e
n
,
p
o
in
t
o
f
in
ter
est
d
o
es
n
o
t
ex
is
t
an
d
th
e
o
b
tain
ed
au
to
co
r
r
elate
d
f
u
n
ctio
n
is
f
la
t
in
n
atu
r
e.
I
f
t
h
e
v
a
lu
e
o
f
λ
1
is
h
i
g
h
an
d
o
t
h
er
λ
2
v
alu
e
is
lo
w
,
t
h
e
n
it
d
en
o
tes
t
h
at
an
ed
g
e
is
f
o
u
n
d
an
d
th
e
co
r
r
esp
o
n
d
in
g
au
to
co
r
r
elate
d
f
u
n
ct
io
n
is
r
ig
id
in
s
h
ap
e.
A
l
s
o
if
th
e
v
alu
e
o
f
b
o
th
λ
1
an
d
λ
2
is
h
i
g
h
t
h
at
m
ea
n
s
a
p
o
in
t
o
f
in
ter
est
is
f
o
u
n
d
an
d
co
r
r
esp
o
n
d
in
g
au
to
co
r
r
elatio
n
f
u
n
ctio
n
is
s
h
ar
p
l
y
p
ick
ed
.
I
t d
en
o
tes th
er
e
is
a
s
i
g
n
i
f
ica
n
t i
n
cr
ea
s
e
i
n
c
(
x
,
y
)
.
3
.
3
.
O
ptic
a
l f
lo
w
m
o
de
l
A
m
o
d
ellin
g
o
f
o
p
tical
f
lo
w
i
s
d
escr
ib
ed
in
th
i
s
s
ec
tio
n
to
an
al
y
s
e
t
h
e
cr
o
w
d
b
eh
a
v
io
u
r
.
I
n
ten
s
it
y
p
ar
am
eter
is
av
o
id
ed
in
esti
m
a
tin
g
o
p
tical
f
lo
w
.
Her
e
w
e
h
a
v
e
u
s
ed
alp
h
a
p
ar
am
eter
s
to
i
m
p
r
o
v
e
th
e
r
esu
lt
an
d
en
h
a
n
ce
t
h
e
p
er
f
o
r
m
a
n
ce
.
A
n
o
r
m
al
m
o
v
e
m
e
n
t
o
f
cr
o
w
d
i
s
s
h
o
w
n
in
F
ig
u
r
e
1
,
w
h
ic
h
co
n
tain
s
a
v
ar
iatio
n
i
n
b
ac
k
g
r
o
u
n
d
.
T
h
is
v
ar
iat
io
n
i
s
s
u
e
co
m
es
i
n
p
atter
n
an
a
l
y
s
is
o
f
th
e
cr
o
w
d
a
n
d
an
al
y
s
in
g
i
n
h
u
m
a
n
d
etec
tio
n
.
I
n
o
r
d
er
to
eli
m
i
n
ate
t
h
is
p
r
o
b
l
e
m
a
n
o
p
tical
f
lo
w
m
o
d
el
i
s
p
r
ep
ar
ed
.
A
s
s
u
m
e
t
h
at
a
n
i
m
ag
e
P
w
it
h
b
ac
k
g
r
o
u
n
d
B
an
d
f
o
r
eg
r
o
u
n
d
F
is
co
n
s
id
er
ed
.
T
h
is
i
m
ag
e
i
s
n
o
t
h
in
g
b
u
t
a
f
r
a
m
e.
T
h
is
f
r
a
m
e
i
s
ac
h
ie
v
ed
f
r
o
m
v
id
eo
s
th
a
t
w
e
g
et
f
r
o
m
s
u
r
v
eilla
n
ce
s
y
s
te
m
an
d
t
h
e
n
u
m
b
er
o
f
f
r
a
m
e
d
e
p
en
d
s
o
n
th
e
v
id
eo
d
u
r
atio
n
.
Fig
u
r
e
1
.
E
x
a
m
p
le
o
f
v
id
eo
f
r
a
m
es i
n
UM
N
d
ataset
A
t
a
n
y
t
i
m
e
t
t
h
e
p
o
s
itio
n
o
f
i
m
ag
e
ca
n
b
e
r
ep
r
esen
ted
as
I
(
p
,
q
,
t)
w
h
er
e
p
an
d
q
is
th
e
co
r
r
esp
o
n
d
in
g
co
o
r
d
in
ates
an
d
th
e
i
m
ag
e
m
o
d
el
is
r
ep
r
esen
ted
as:
ℱ
(
,
,
)
.
(
,
,
)
+
ℬ
(
,
,
)
.
(
1
−
(
,
,
)
)
=
(
,
,
)
(
8
)
T
h
is
m
o
d
el
is
u
s
e
f
u
l
in
esti
m
a
tin
g
th
e
p
ar
am
e
ter
s
o
f
th
e
s
eq
u
en
t
ial
i
m
a
g
e.
W
e
ass
u
m
e
t
h
at
th
e
i
m
ag
e
in
ter
v
a
l
is
2
5
f
r
a
m
es
p
er
s
ec
o
n
d
i.e
.
(
t
ϵ
)
.
T
o
ac
h
iev
e
th
i
s
m
o
d
el,
it
is
co
n
s
id
er
ed
th
at
t
h
e
b
ac
k
g
r
o
u
n
d
i
s
s
tatio
n
ar
y
.
Mo
v
ab
le
o
b
j
ec
ts
ar
e
r
ep
r
esen
ted
b
y
h
i
g
h
er
b
r
ig
h
t
n
es
s
.
B
ac
k
g
r
o
u
n
d
p
i
x
el
i
s
est
i
m
ated
b
y
th
e
lo
w
e
s
t
i
m
a
g
e
b
r
ig
h
tn
e
s
s
.
T
h
e
v
al
u
e
o
f
ex
tr
ac
ted
p
ix
el
ca
n
b
e
d
ef
in
e
d
as:
1
(
,
)
=
min
(
(
,
,
1
)
,
(
,
,
2
)
…
.
,
(
,
,
)
)
(
9
)
Her
e
b1
is
ex
tr
ac
ted
p
ix
el
an
d
t
h
i
s
is
n
o
t
d
ep
en
d
en
t
o
n
t
h
e
ti
m
e
in
ter
v
a
l.
Usi
n
g
t
h
e
v
al
u
e
o
f
b1
w
e
ca
n
esti
m
ate
th
e
f
o
r
eg
r
o
u
n
d
i
m
ag
e
f1
th
at
ca
n
b
e
d
ef
in
ed
as:
1
(
,
,
)
=
(
,
,
)
−
1
(
,
)
(
1
0
)
I
f
th
e
n
u
m
b
er
o
f
f
r
a
m
e
s
v
ar
ie
s
,
th
e
in
te
n
s
i
t
y
o
f
t
h
at
f
r
a
m
e
w
ill
also
v
ar
y
.
I
t
also
ef
f
ec
ts
o
n
t
h
e
i
m
ag
e
b
ac
k
g
r
o
u
n
d
an
d
f
o
r
eg
r
o
u
n
d
.
So
th
e
v
ar
iati
o
n
in
i
n
ten
s
it
y
ca
n
co
m
p
u
ted
i
n
(
11
)
.
=
(
−
)
/
(
ℱ
−
)
(
1
1
)
β
d
en
o
tes in
te
n
s
i
t
y
tr
a
n
s
f
o
r
m
ati
o
n
.
No
w
w
e
n
ee
d
to
co
m
p
u
te
t
h
e
o
p
tical
f
lo
w
o
f
th
e
h
u
m
a
n
o
r
an
y
m
o
v
in
g
o
b
j
ec
t
in
th
e
cr
o
w
d
s
ce
n
e.
I
n
o
u
r
w
o
r
k
,
w
e
ass
u
m
e
t
h
at
th
e
t
o
tal
s
p
ac
e
d
er
iv
ativ
e
an
d
ti
m
e
d
er
iv
ativ
e
ar
e
s
to
r
ed
in
β
an
d
ea
ch
o
f
th
e
p
ar
ticle
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
C
r
o
w
d
r
ec
o
g
n
itio
n
s
ystem
b
a
s
ed
o
n
o
p
tica
l flo
w
a
lo
n
g
w
ith
S
V
M c
la
s
s
ifier
(
S
h
r
ee
d
a
r
s
h
a
n
.
K
)
2455
in
f
r
a
m
e
i
s
d
en
o
ted
b
y
s
.
Fu
r
th
e
r
w
e
f
o
llo
w
T
a
y
lo
r
ex
p
a
n
s
i
o
n
o
n
th
e
i
n
p
u
t
f
r
a
m
e
s
eq
u
en
ce
a
n
d
i
m
p
le
m
e
n
ti
n
g
p
o
s
itio
n
o
f
th
e
i
m
a
g
e
p
,
q
,
t
.
I
t c
an
b
e
g
iv
e
n
as:
ℰ
(
)
≈
(
)
+
(
)
+
(
1
2
)
p
d
en
o
tes
f
ir
s
t
o
r
d
er
d
e
r
iv
ativ
e
s
w
i
th
r
esp
ec
t
to
ti
m
e
an
d
s
p
ac
e
d
ata
v
ec
to
r
.
T
h
e
f
u
n
ct
io
n
o
f
o
p
tical
f
lo
w
esti
m
atio
n
i
s
g
i
v
e
n
as e
q
u
atio
n
.
T
h
e
f
u
n
ctio
n
o
f
o
p
tical
f
lo
w
esti
m
a
tio
n
is
g
i
v
en
a
s
eq
u
atio
n
.
(
,
)
=
∑
(
)
2
(
,
)
(
1
3
)
A
co
m
p
lete
f
lo
w
d
iag
r
a
m
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
is
s
h
o
w
n
i
n
F
i
g
u
r
e
2
.
Fig
u
r
e
2
.
C
o
m
p
lete
f
lo
w
d
ia
g
r
a
m
o
f
p
r
o
p
o
s
ed
m
o
d
el
4.
SI
M
UL
AT
I
O
N
R
E
S
UL
T
A
ND
ANAL
YSI
S
T
h
is
s
ec
tio
n
d
e
s
cr
ib
es
th
e
r
esu
lt
esti
m
atio
n
a
n
d
th
e
p
r
o
p
o
s
ed
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
T
h
is
m
o
d
el
is
i
m
p
le
m
en
ted
w
it
h
t
h
e
h
elp
o
f
Ma
t
L
ab
u
s
er
in
ter
f
ac
e.
T
h
e
v
i
d
eo
d
ata
s
et
w
h
at
w
e
h
av
e
u
s
e
d
in
o
u
r
ap
p
r
o
ac
h
is
tak
en
f
r
o
m
Un
i
v
er
s
i
t
y
o
f
Mi
n
n
e
s
o
ta
an
d
co
n
tain
n
o
r
m
a
l
o
r
ab
n
o
r
m
al
ac
tiv
itie
s
.
T
o
ev
alu
ate
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
,
w
e
h
a
v
e
u
s
ed
G
R
OUND
tr
u
t
h
o
f
d
ataset.
I
n
te
r
m
ed
iate
r
es
u
lt
o
f
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
is
s
h
o
w
n
in
th
e
f
o
r
m
o
p
tical
f
lo
w
a
n
d
s
t
r
ea
k
lin
e
f
lo
w
.
First
th
e
v
id
eo
s
eq
u
en
ce
is
s
el
ec
ted
an
d
co
n
v
er
ted
in
to
f
r
a
m
es.
No
w
ea
ch
f
r
a
m
e
is
ta
k
e
n
as
an
in
p
u
t
o
n
a
s
p
ec
if
ic
ti
m
e
in
ter
v
a
l,
th
en
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
s
is
ap
p
lied
.
I
n
p
r
e
-
p
r
o
ce
s
s
in
g
s
t
ep
s
im
a
g
e
f
r
a
m
e
s
is
co
n
v
er
ted
in
to
g
r
a
y
s
ca
le
an
d
f
u
r
t
h
er
n
o
is
e
r
e
m
o
v
i
n
g
a
n
d
en
h
a
n
ci
n
g
p
r
o
ce
s
s
is
p
er
f
o
r
m
e
d
.
T
h
e
ex
p
er
im
e
n
tal
re
s
u
lt
s
ar
e
g
iv
e
n
i
n
th
e
f
o
r
m
o
f
F
i
g
u
r
e
s
3
,
4
,
5
an
d
6
.
Fig
u
r
e
3
.
o
p
tical
f
lo
w
o
f
th
e
v
id
eo
s
eq
u
e
n
ce
Fig
u
r
e
4
.
Stre
ak
lin
e
f
lo
w
o
f
th
e
v
id
eo
s
eq
u
en
ce
Fig
u
r
e
5
.
No
r
m
al
f
r
a
m
e
s
Fig
u
r
e
6
.
A
b
n
o
r
m
al
f
r
a
m
e
s
I
n
Fig
u
r
e
.
3
an
o
p
tical
f
lo
w
p
a
tter
n
d
iag
r
a
m
is
s
h
o
w
n
.
Fig
u
r
e
4
illu
s
tr
ates
t
h
e
s
tr
ea
k
li
n
e
f
lo
w
o
f
t
h
e
co
r
r
esp
o
n
d
in
g
f
r
a
m
e.
Ne
x
t
in
Fig
u
r
e
5
s
o
m
e
p
r
o
ce
s
s
in
g
h
a
p
p
en
s
th
at
b
elo
n
g
s
to
a
n
al
y
s
i
n
g
f
r
a
m
es.
Fi
g
u
r
e
6
d
em
o
n
s
tr
ate
s
th
a
t th
e
ab
n
o
r
m
a
l a
ctiv
it
y
is
s
tar
tin
g
n
o
w
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
87
08
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t 2
0
1
9
:
2
4
5
1
-
2459
2456
Nex
t,
w
e
ar
e
j
u
s
t
co
m
p
ar
i
n
g
o
u
r
p
r
o
p
o
s
ed
alg
o
r
ith
m
to
g
r
o
u
n
d
tr
u
th
a
n
d
SF
M
m
o
d
el.
I
n
o
u
r
m
o
d
el
w
e
to
o
k
4
2
f
r
a
m
es
o
f
cr
o
w
d
s
a
m
p
le,
4
4
f
r
a
m
es
o
f
co
u
r
t
y
ar
d
s
a
m
p
le
a
n
d
3
6
f
r
a
m
e
s
o
f
co
r
r
id
o
r
s
am
p
le.
I
n
Fi
g
u
r
e
s
7
,
8
an
d
9
,
w
e
h
av
e
s
h
o
w
n
t
h
e
co
m
p
ar
ati
v
e
r
esu
lt
o
f
o
u
r
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
to
g
r
o
u
n
d
tr
u
th
a
n
d
SF
M
m
o
d
el
ac
r
o
s
s
f
r
a
m
e
s
.
Fig
u
r
e
7
.
C
o
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
alg
o
r
ith
m
w
it
h
Gr
o
u
n
d
tr
u
th
a
n
d
ex
i
s
ti
n
g
al
g
o
r
ith
m
(
S
F
M)
f
o
r
d
ataset
cr
o
w
d
.
Fig
u
r
e
8
. C
o
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
alg
o
r
ith
m
w
it
h
Gr
o
u
n
d
tr
u
th
a
n
d
ex
i
s
ti
n
g
al
g
o
r
ith
m
(
S
F
M)
f
o
r
d
ataset
co
u
r
t
y
ar
d
Fig
u
r
e
8
.
C
o
m
p
ar
is
o
n
o
f
p
r
o
p
o
s
ed
alg
o
r
ith
m
w
it
h
Gr
o
u
n
d
tr
u
th
a
n
d
ex
i
s
ti
n
g
al
g
o
r
ith
m
(
S
F
M)
f
o
r
d
ataset
co
r
r
id
o
r
As
f
r
o
m
th
e
ab
o
v
e
g
r
ap
h
s
a
n
d
it
is
p
r
o
v
ed
th
at
o
u
r
ap
p
r
o
ac
h
g
i
v
es
b
etter
r
es
u
lt
co
m
p
ar
e
to
ex
is
ti
n
g
(
SF
M)
alg
o
r
ith
m.
R
OC
is
a
m
o
s
t
co
m
m
o
n
tech
n
iq
u
e
to
v
is
u
alize
p
er
f
o
r
m
a
n
ce
o
f
a
class
if
i
er
.
Her
e,
w
e
h
a
v
e
u
s
ed
SVM
clas
s
i
f
ier
w
h
ic
h
is
b
in
ar
y
.
B
in
ar
y
i
n
th
e
s
e
n
s
e,
th
at
w
e
h
a
v
e
o
n
l
y
t
w
o
clas
s
.
A
c
co
r
d
in
g
to
o
u
r
p
ap
e
r
s
ce
n
ar
io
a
cr
o
w
d
ca
n
b
e
class
i
f
ied
in
to
t
w
o
ca
te
g
o
r
ies
f
ir
s
t
o
n
e
is
n
o
r
m
al
an
d
o
th
er
i
s
ab
n
o
r
m
al.
T
h
ese
SVM
class
i
f
ier
h
as
a
s
u
cc
e
s
s
r
ate
o
f
9
7
%
w
h
ic
h
is
b
etter
co
m
p
ar
is
o
n
to
ea
r
lier
SF
M
m
et
h
o
d
.
R
OC
cu
r
v
e
is
s
h
o
w
n
in
Fig
u
r
e
1
0
,
1
1
an
d
1
2
f
o
r
d
if
f
er
en
t
UM
N
d
ata
s
et.
T
ab
le
1
in
d
icate
s
th
a
t
t
h
e
p
r
o
p
o
s
ed
m
et
h
o
d
h
as
a
b
etter
s
u
cc
e
s
s
r
ate
w
h
en
co
m
p
ar
ed
to
th
e
o
th
er
m
e
th
o
d
s
.
Fig
u
r
e
9
.
R
o
c
cu
r
v
e
f
o
r
SVM
class
i
f
ier
f
o
r
cr
o
w
d
(
9
7
.
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
C
r
o
w
d
r
ec
o
g
n
itio
n
s
ystem
b
a
s
ed
o
n
o
p
tica
l flo
w
a
lo
n
g
w
ith
S
V
M c
la
s
s
ifier
(
S
h
r
ee
d
a
r
s
h
a
n
.
K
)
2457
Fig
u
r
e
10
.
R
o
c
cu
r
v
e
f
o
r
SVM
class
i
f
ier
f
o
r
co
u
r
t
y
ar
d
(
9
7
.
3
6
%)
Fig
u
r
e
11
.
R
o
c
cu
r
v
e
f
o
r
SVM
class
i
f
ier
f
o
r
co
r
r
id
o
r
(
9
5
.
6
5
%)
T
ab
le
1
.
C
o
m
p
ar
is
o
n
w
i
th
d
i
f
f
er
en
t
s
tate
o
f
ar
t
s
m
et
h
o
d
AU
C
M
e
t
h
o
d
A
U
C
S
F
M
[
3
3
]
9
4
.
9
C
h
a
o
t
i
c
i
n
v
a
r
i
a
n
t
s
[
3
4
]
9
9
.
4
S
p
a
r
se
r
e
c
o
n
s.
[
3
5
]
9
9
.
6
L
o
c
a
l
st
a
t
i
st
i
c
s
[
3
6
]
9
9
.
5
M
D
T
[
3
7
]
9
9
.
5
H
O
G
+
H
O
S
[
3
8
]
9
7
.
0
2
L
M
V
D
[
3
9
]
9
8
.
6
1
O
U
R
P
R
O
P
O
S
ED
9
9
.
7
1
5.
CO
NCLU
SI
O
N
I
n
o
r
d
er
to
in
cr
ea
s
e
th
e
ac
cu
r
ac
y
o
f
th
e
b
eh
a
v
io
u
r
o
f
an
ab
n
o
r
m
al
cr
o
w
d
d
etec
tio
n
s
y
s
te
m
we
p
r
o
p
o
s
ed
o
p
tical
f
lo
w
m
et
h
o
d
.
I
n
itiall
y
w
e
h
a
v
e
d
o
n
e
p
r
e
-
p
r
o
ce
s
s
in
g
o
n
th
e
co
n
v
er
ti
n
g
i
m
a
g
e
f
r
a
m
es
e
x
tr
ac
ted
f
r
o
m
v
id
eo
s
a
m
p
le.
T
h
is
i
m
a
g
e
f
r
a
m
e
i
s
s
u
b
j
ec
ted
to
m
o
tio
n
h
ea
t
m
ap
,
w
h
ic
h
ex
tr
ac
ts
t
h
e
m
ai
n
r
eg
io
n
o
f
m
o
tio
n
.
L
ater
p
o
in
ts
o
f
in
ter
es
t
ar
e
ex
tr
ac
ted
f
r
o
m
t
h
o
s
e
ar
ea
s
o
f
m
o
ti
o
n
h
ea
t
m
ap
.
T
h
is
p
r
o
ce
s
s
r
ed
u
ce
s
o
u
r
ca
lcu
latio
n
co
m
p
le
x
it
y
.
O
n
ce
w
e
g
e
t
th
e
p
o
in
t
o
f
in
ter
est,
w
e
co
m
p
u
te
o
p
tical
f
lo
w
m
o
d
el.
Op
tical
f
lo
w
m
o
d
el
d
escr
ib
es
th
e
f
lo
w
o
f
m
o
tio
n
o
f
t
h
e
cr
o
w
d
.
Af
ter
an
a
l
y
s
i
n
g
th
e
o
p
tical
f
l
o
w
m
o
d
el,
w
e
s
et
a
t
h
r
es
h
o
ld
p
o
in
t.
T
h
is
t
h
r
es
h
o
ld
v
alu
e
i
s
s
u
b
j
ec
ted
to
SVM
class
if
ier
.
I
f
t
h
e
v
a
lu
e
o
f
t
h
e
o
p
tical
f
lo
w
i
s
les
s
th
a
n
t
h
e
th
r
es
h
o
ld
v
alu
e,
t
h
e
clas
s
i
f
ier
p
ass
es a
d
ec
is
io
n
n
o
r
m
al.
I
f
th
e
o
p
tical
f
lo
w
v
al
u
e
is
g
r
ea
ter
th
an
t
h
e
th
r
es
h
o
ld
v
alu
e,
t
h
en
th
e
s
y
s
te
m
p
as
s
es a
d
ec
is
io
n
ab
n
o
r
m
al.
T
h
is
cla
s
s
i
f
icatio
n
h
as 9
9
.
7
1
% a
cc
u
r
ac
y
.
RE
F
E
R
E
NC
E
S
[1
]
X
.
S
.
L
i
,
e
t
a
l.
,
"
A
n
a
l
y
sis
a
n
d
S
i
m
p
li
f
ica
ti
o
n
o
f
T
h
re
e
-
Di
m
e
n
sio
n
a
l
S
p
a
c
e
V
e
c
to
r
P
W
M
f
o
r
T
h
re
e
-
P
h
a
se
F
o
u
r
-
L
e
g
In
v
e
rters
,
"
IEE
E
T
ra
n
s
a
c
ti
o
n
s o
n
In
d
u
stria
l
El
e
c
tro
n
ics
,
v
o
l
.
5
8
,
p
p
.
4
5
0
-
4
6
4
,
F
e
b
2
0
1
1
.
[2
]
K.
k
ra
tzL
n
ish
in
o
,
“
A
n
o
m
a
l
y
d
e
tec
ti
o
n
i
n
e
x
trem
e
l
y
c
ro
w
d
e
d
sc
e
n
e
s
sp
a
ti
o
-
t
e
m
p
o
ra
l
m
o
ti
o
n
p
a
tt
e
rn
m
o
d
e
ls,”
in
:
Pr
o
c
.
CVP
R
,
p
p
.
1
4
4
6
-
1
4
5
3
,
2
0
0
9
.
[3
]
J.
Xu
,
S
.
De
n
m
a
n
,
C.
F
o
o
k
e
s
e
t.
a
l,
“
Un
u
su
a
l
e
v
e
n
t
d
e
tec
ti
o
n
i
n
c
ro
w
d
e
d
sc
e
n
e
s
u
si
n
g
b
a
g
o
f
L
BP
s
i
n
s
p
a
ti
o
-
tem
p
o
ra
l
p
a
tch
e
s,”
in
:
Pro
c
.
DICTA
,
p
p
.
5
4
9
-
5
5
4
,
2
0
1
1
.
[4
]
E.
L
.
A
n
d
ra
d
e
,
S
.
Blu
n
sd
e
n
,
a
n
d
R.
B.
F
ish
e
r,
“
Hid
d
e
n
m
a
rk
o
v
m
o
d
e
ls
f
o
r
o
p
ti
c
a
l
f
lo
w
a
n
a
l
y
sis
in
c
r
o
w
d
s,”
in
:
Pro
c
.
ICPR
,
p
p
.
4
6
0
-
4
6
3
,
2
0
0
6
.
[5
]
G
u
o
Jin
g
*
,
Zh
o
u
M
i
n
g
q
u
a
n
,
L
i
Ch
a
o
,
“
A
u
to
m
a
ti
c
3
D
M
o
d
e
l
A
n
n
o
tati
o
n
b
y
a
Tw
o
-
Di
m
e
n
sio
n
a
l
Hid
d
e
n
M
a
rk
o
v
M
o
d
e
l
,”
T
EL
KOM
NIKA
I
n
d
o
n
e
sia
n
J
o
u
rn
a
l
o
f
El
e
c
trica
l
E
n
g
in
e
e
rin
g
,
V
o
l.
1
2
,
No
.
5
,
p
p
.
3
2
7
2
~
3
2
8
0
DO
I:
h
tt
p
:
//
d
x
.
d
o
i.
o
rg
/1
0
.
1
1
5
9
1
/t
e
lk
o
m
n
ik
a
.
v
1
2
i5
.
4
9
4
6
,
M
a
y
2
0
1
4
.
[6
]
S
.
A
li
a
n
d
M
.
S
h
a
h
,
“
A
l
a
g
r
a
n
g
ia
n
p
a
rticle
d
y
n
a
mic
s
a
p
p
ro
a
c
h
f
o
r
c
ro
wd
f
lo
w
se
g
me
n
ta
ti
o
n
a
n
d
sta
b
il
it
y
a
n
a
lys
is
,
”
in
:
Pr
o
c
.
CVP
R
,
p
p
.
1
-
6
,
2
0
0
7
.
[7
]
R.
M
e
h
ra
n
,
A
.
Oy
a
m
a
,
a
n
d
M
.
S
h
a
h
,
“
Ab
n
o
rm
a
l
c
ro
w
d
b
e
h
a
v
io
r
d
e
t
e
c
ti
o
n
u
sin
g
so
c
i
a
l
f
o
rc
e
mo
d
e
l
,
”
i
n
:
Pro
c
.
CVP
R
,
p
p
.
9
3
5
-
9
4
2
,
2
0
0
9
.
[8
]
M
e
n
g
L
i
*
1
,
J
ia
-
h
o
n
g
L
ian
g
1
,
S
h
i
-
lei
L
i
2
,
“
S
o
ld
iers
’
G
ro
u
p
Be
h
a
v
io
rs
S
i
m
u
latio
n
Ba
se
d
o
n
Im
p
ro
v
e
d
M
A
P
RM
,
”
T
e
lec
o
mm
u
n
ica
ti
o
n
C
o
mp
u
ti
n
g
El
e
c
tro
n
ics
a
n
d
Co
n
tro
l
(
T
E
L
KOM
NIKA
)
,
V
o
l.
1
1
,
No
.
5
,
p
p
.
2
5
4
5
~
2
5
5
2
IS
S
N
:
2
3
0
2
-
4
0
4
6
,
M
a
y
2
0
1
3
.
[9
]
S
.
W
u
,
B.
E.
M
o
o
re
a
n
d
M
.
S
h
a
h
,
“
Ch
a
o
ti
c
in
v
a
ri
a
n
ts
o
f
la
g
ra
n
g
ia
n
p
a
rt
icle
tra
jec
to
rie
s
fo
r
a
n
o
ma
l
y
d
e
tec
ti
o
n
in
c
ro
wd
e
d
sc
e
n
e
s
,
”
in
:
P
r
o
c
.
C
VPR,
p
p
.
2
0
5
4
-
2
0
6
0
,
2
0
1
0
.
[1
0
]
S
.
A
li
a
n
d
M
.
S
h
a
h
,
“
H
u
m
a
n
a
c
ti
o
n
re
c
o
g
n
it
i
o
n
in
v
id
e
o
s
u
sin
g
k
i
n
e
m
a
ti
c
fe
a
tu
re
s
a
n
d
m
u
lt
ip
le
in
s
tan
c
e
lea
rn
in
g
,
”
IEE
E
T
ra
n
s.
P
a
tt
.
An
a
l.
M
a
c
h
.
I
n
tell
.
,
v
o
l.
3
2
,
p
p
.
2
8
8
-
3
0
3
,
2
0
1
0
.
[1
1
]
T
.
Ha
s
s
n
e
r,
Y.
Itch
e
r
a
n
d
O.
Klip
e
r
-
G
ro
ss
,
“
Vi
o
len
t
fl
o
ws: R
e
a
l
-
ti
me
d
e
tec
ti
o
n
o
f
v
io
len
t
c
ro
w
d
b
e
h
a
v
io
r
,
”
in
:
Pro
c
.
CVP
RW
,
p
p
.
1
-
6
,
2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
87
08
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
9
,
No
.
4
,
A
u
g
u
s
t 2
0
1
9
:
2
4
5
1
-
2459
2458
[1
2
]
J.
Kim
a
n
d
K.
G
ra
u
m
a
n
,
“
Ob
se
rv
e
lo
c
a
ll
y
,
i
n
fer
g
lo
b
a
ll
y
:
A
s
p
a
c
e
-
ti
me
M
RF
f
o
r
d
e
tec
ti
n
g
a
b
n
o
rm
a
l
a
c
ti
v
it
ies
w
it
h
in
c
re
me
n
ta
l
u
p
d
a
tes
,
”
i
n
Pro
c
.
I
E
EE
Co
n
f.
CV
PR
,
p
p
.
2
9
2
1
–
2
9
2
8
,
Ju
n
.
2
0
0
9
.
[1
3
]
Y.
Be
n
e
z
e
th
,
P
.
-
M
.
J
o
d
o
in
,
V
.
S
a
l
ig
ra
m
a
,
a
n
d
C.
Ro
se
n
b
e
rg
e
r,
“
A
b
n
o
rm
a
l
e
v
e
n
ts
d
e
tec
ti
o
n
b
a
se
d
o
n
sp
a
ti
o
-
tem
p
o
ra
l
co
-
o
c
c
u
re
n
c
e
s,”
in
C
o
mp
u
ter
Vi
si
o
n
a
n
d
P
a
tt
e
rn
Rec
o
g
n
it
io
n
(
CVP
R),
p
p
.
2
4
5
8
–
2
4
6
5
,
2
0
0
9
.
[1
4
]
W
e
n
x
i
L
iu
,
Ry
n
so
n
W
.
H.
L
a
u
a
n
d
Din
e
sh
M
a
n
o
c
h
a
“
Ro
b
u
st
i
n
d
iv
id
u
a
l
a
n
d
h
o
li
st
ic
f
e
a
tu
re
s
f
o
r
c
ro
w
d
sc
e
n
e
c
las
si
f
ica
ti
o
n
.
”
Pa
tt
e
r
n
Rec
o
g
n
it
i
o
n
5
8
,
1
1
0
-
1
2
0
,
2
0
1
6
.
[1
5
]
A
.
Be
ra
,
S
.
Ki
m
a
n
d
D.
M
a
n
o
c
h
a
,
"
Rea
lt
ime
An
o
m
a
ly
De
tec
ti
o
n
Us
in
g
T
ra
jec
to
ry
-
L
e
v
e
l
Cro
wd
Be
h
a
v
io
r
L
e
a
rn
i
n
g
,
"
2
0
1
6
IE
EE
Co
n
fer
e
n
c
e
o
n
C
o
m
p
u
ter
Vi
si
o
n
a
n
d
P
a
tt
e
rn
Rec
o
g
n
it
io
n
W
o
rk
sh
o
p
s
(
CVP
RW
)
,
L
a
s
V
e
g
a
s,
NV
,
p
p
.
1
2
8
9
-
1
2
9
6
,
2
0
1
6
.
[1
6
]
Na
j
m
u
z
z
a
m
a
Zerd
i,
S
u
b
h
a
sh
S
Ku
lk
a
rn
i,
V
D
M
y
tri
,
Ka
sh
y
a
p
D
Dh
ru
v
e
,
”
Cro
w
d
Be
h
a
v
io
r
A
n
a
l
y
sis
a
n
d
Clas
sif
ic
a
ti
o
n
u
si
n
g
G
ra
p
h
T
h
e
o
re
ti
c
A
p
p
ro
a
c
h
,”
Glo
b
a
l
J
o
u
rn
a
l
o
f
Co
mp
u
ter
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
2
8
M
a
rc
h
;
V
o
l
u
m
e
1
4
Iss
u
e
1
V
e
rsio
n
1
.
0
.
I
S
S
N:
0
9
7
5
-
4
1
7
2
,
2
0
1
4
.
[1
7
]
O.
P
.
P
o
p
o
o
la
a
n
d
K.
W
a
n
g
,
“
Vid
e
o
-
b
a
se
d
a
b
n
o
rm
a
l
h
u
m
a
n
b
e
h
a
v
io
r
re
c
o
g
n
it
io
n
-
A
re
v
ie
w
,
”
IE
EE
T
ra
n
s.
S
y
st
.
,
M
a
n
,
Cy
b
e
rn
.
C,
A
p
p
l.
Re
v
.
,
v
o
l.
4
2
,
n
o
.
6
,
p
p
.
8
6
5
–
8
7
8
,
N
o
v
.
2
0
1
2
.
[1
8
]
A
.
A
d
a
m
,
E.
Riv
li
n
,
I.
S
h
im
sh
o
n
i
,
a
n
d
D.
Re
in
it
z
,
“
Ro
b
u
st
re
a
l
-
ti
m
e
u
n
u
s
u
a
l
e
v
e
n
t
d
e
tec
ti
o
n
u
si
n
g
m
u
lt
ip
le
f
ix
e
d
-
lo
c
a
ti
o
n
m
o
n
it
o
rs,”
IE
EE
T
r
a
n
s.
Pa
tt
e
rn
A
n
a
l.
M
a
c
h
.
In
tell
.
,
v
o
l.
3
0
,
n
o
.
3
,
p
p
.
5
5
5
–
5
6
0
,
M
a
r.
2
0
0
8
.
[1
9
]
Z.
Ju
n
a
n
d
L
.
Zh
ij
in
g
,
“
De
tec
ti
n
g
irre
g
u
la
rit
ies
b
y
ima
g
e
c
o
n
t
o
u
r
b
a
se
d
o
n
f
u
zz
y
n
e
u
ra
l
n
e
two
rk
,
”
i
n
Pro
c
.
3
rd
In
t
.
Co
n
f.
In
n
o
v
a
ti
v
e
Co
mp
u
t.
I
n
f.
C
o
n
tro
l
,
Ju
n
.
1
8
–
2
0
,
2
0
0
8
,
p
p
.
4
0
1
.
[2
0
]
J.
Kim
a
n
d
K.
G
ra
u
m
a
n
,
“
Ob
se
rv
e
lo
c
a
ll
y
,
i
n
fer
g
lo
b
a
ll
y
:
A
s
p
a
c
e
-
ti
me
M
RF
f
o
r
d
e
tec
ti
n
g
a
b
n
o
rm
a
l
a
c
ti
v
it
ies
w
it
h
in
c
re
me
n
ta
l
u
p
d
a
tes
,
”
i
n
Pro
c
.
I
E
EE
Co
n
f.
C
o
mp
u
t.
V
isio
n
Pa
tt
e
rn
Rec
o
g
,
p
p
.
2
9
2
1
–
2
9
2
8
,
2
0
0
9
.
[2
1
]
L
.
Kra
tz
a
n
d
K.
Nish
i
n
o
,
“
An
o
m
a
ly
d
e
tec
ti
o
n
in
e
x
tre
me
ly
c
ro
wd
e
d
sc
e
n
e
s
u
si
n
g
sp
a
ti
o
-
tem
p
o
ra
l
mo
ti
o
n
p
a
t
ter
n
mo
d
e
ls,”
i
n
Pr
o
c
.
IEE
E
Co
n
f.
C
o
mp
u
t.
Vi
sio
n
P
a
tt
e
rn
Rec
o
g
,
p
p
.
1
4
4
6
–
1
4
5
3
,
2
0
0
9
.
[2
2
]
T
ian
W
a
n
g
,
a
n
d
Hic
h
e
m
S
n
o
u
ss
i,
“
De
tec
ti
o
n
o
f
A
b
n
o
rm
a
l
V
isu
a
l
Ev
e
n
ts
v
ia
G
lo
b
a
l
Op
ti
c
a
l
F
lo
w
Orie
n
tatio
n
Histo
g
ra
m
”
IEE
E
T
ra
n
s.
In
f
o
rm
a
ti
o
n
Fo
re
n
sic
s a
n
d
S
e
c
u
rity
,
v
o
l.
9
,
n
o
.
6
,
p
p
:
9
8
8
-
9
9
8
,
J
u
n
.
2
0
1
4
.
[2
3
]
S
.
Kw
a
k
a
n
d
H.
By
u
n
,
“
De
tec
ti
o
n
o
f
d
o
m
in
a
n
t
f
lo
w
a
n
d
a
b
n
o
rm
a
l
e
v
e
n
ts
in
s
u
rv
e
il
lan
c
e
v
id
e
o
,
”
Op
ti
c
a
l
E
n
g
i
n
e
e
rin
g
,
v
o
l.
5
0
,
n
o
.
2
,
p
p
.
0
2
7
2
0
2
-
1
–
0
2
7
2
0
2
-
8
,
2
0
1
1
.
[2
4
]
Y.
Zh
o
u
,
S
.
Ya
n
,
a
n
d
T
.
S
.
H
u
a
n
g
,
“
De
tec
ti
n
g
a
n
o
m
a
ly i
n
v
id
e
o
s
fr
o
m t
r
a
jec
to
ry
simil
a
rity a
n
a
lys
is
,
”
in
Pro
c
.
IEE
E
In
t.
C
o
n
f
.
M
u
lt
ime
d
ia
Ex
p
o
,
p
p
.
1
0
8
7
–
1
0
9
0
,
2
0
0
7
.
[2
5
]
M
.
T
h
id
a
,
H.
En
g
,
M
.
D
o
ro
t
h
y
,
a
n
d
P
.
P
e
m
a
g
n
in
o
,
“
L
e
a
rn
i
n
g
v
i
d
e
o
ma
n
if
o
ld
fo
r
se
g
me
n
ti
n
g
c
ro
wd
e
v
e
n
ts
a
n
d
a
b
n
o
rm
a
li
ty d
e
tec
ti
o
n
,
”
i
n
Pro
c
.
Asia
n
C
o
n
f.
Co
m
p
u
t
.
Vi
si
on
,
p
p
.
4
3
9
–
4
4
9
,
2
0
1
0
.
[2
6
]
S
i
W
u
,
Ha
u
-
S
a
n
W
o
n
g
,
a
n
d
Zh
i
w
e
n
Yu
,
“
A
Ba
y
e
sia
n
M
o
d
e
l
f
o
r
Cro
w
d
Esc
a
p
e
Be
h
a
v
io
r
De
tec
ti
o
n
,
”
IEE
E
T
ra
n
s
.
Circ
u
it
s a
n
d
S
y
ste
ms
fo
r V
i
d
e
o
T
e
c
h
n
o
l
o
g
y
,
v
o
l.
2
4
,
n
o
.
1
,
p
p
:
8
5
-
98
,
Ja
n
.
2
0
1
4
.
[2
7
]
V
.
S
a
li
g
ra
m
a
a
n
d
Z.
C
h
e
n
,
“
Vi
d
e
o
a
n
o
ma
ly
d
e
tec
ti
o
n
b
a
se
d
o
n
l
o
c
a
l
sta
ti
stica
l
a
g
g
re
g
a
tes
,
”
i
n
C
o
mp
u
ter
Vi
sio
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
(
CVP
R),
2
0
1
2
IEE
E
Co
n
fer
e
n
c
e
o
n
.
IEE
E
,
p
p
.
2
1
1
2
–
2
1
1
9
,
2
0
1
2
.
[2
8
]
Ha
n
g
S
u
,
Hu
a
Ya
n
g
,
S
h
i
b
a
o
Zh
e
n
g
,
Ya
we
n
F
a
n
,
a
n
d
S
h
a
W
e
i,
“
T
h
e
L
a
rg
e
-
S
c
a
le Cro
w
d
Be
h
a
v
io
r
P
e
rc
e
p
ti
o
n
Ba
se
d
o
n
S
p
a
ti
o
-
T
e
m
p
o
ra
l
Visc
o
u
s
F
lu
i
d
F
iel
d
,
”
IE
EE
T
r
a
n
s.
In
f
o
rm
a
ti
o
n
F
o
re
n
sic
s a
n
d
S
e
c
u
rity
,
v
o
l.
8
,
n
o
.
1
0
,
p
p
:1
5
7
5
-
1
5
8
9
,
Oc
t.
2
0
1
3
.
[2
9
]
V
.
D.
M
y
tri
a
n
d
Ka
sh
y
a
p
D
Dh
r
u
v
e
N
Zerd
i,
S
S
Ku
lk
a
rn
i,
”
Cr
o
w
d
b
e
h
a
v
io
u
r
a
n
a
y
sis
c
o
n
sid
e
rin
g
in
ter
-
p
e
rso
n
n
e
l
a
c
ti
v
it
ies
in
su
rv
il
la
n
c
e
sy
ste
m
s
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
of
Co
m
p
u
t
e
r
En
g
i
n
e
e
rin
g
&
T
e
c
h
n
o
lo
g
y
,
F
e
b
ru
a
ry
;
V
o
l
5
,
No
.
2
,
P
a
g
e
s 7
1
-
87
,
2
0
1
4
.
[3
0
]
Y.
Be
n
e
z
e
th
,
P
.
-
M
.
Jo
d
o
in
,
a
n
d
V.
S
a
li
g
ra
m
a
,
“
A
b
n
o
rm
a
li
t
y
d
e
tec
ti
o
n
u
sin
g
l
o
w
-
lev
e
l
c
o
-
o
c
c
u
rrin
g
e
v
e
n
ts,”
Pa
tt
e
rn
Rec
o
g
n
it
io
n
L
e
tt
e
r,
v
o
l
.
3
2
,
n
o
.
3
,
p
p
.
4
2
3
–
4
3
1
,
2
0
1
1
.
[3
1
]
Ya
sir
S
a
li
h
A
li
,
M
o
h
a
m
m
e
d
T
a
l
a
l
S
im
si
m
,
”
V
isu
a
l
S
u
rv
e
il
lan
c
e
f
o
r
Ha
jj
a
n
d
Um
ra
h
:
A
Re
v
ie
w
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Arti
fi
c
i
a
l
I
n
telli
g
e
n
c
e
(
IJ
-
AI)
;
Vo
l.
3
,
No
.
2
,
ju
n
e
2
0
1
4
.
[3
2
]
A
li
Ba
z
m
i,
k
a
ri
m
F
a
e
z
,
”
In
c
re
a
sin
g
th
e
Ac
c
u
ra
c
y
o
f
De
t
e
c
ti
o
n
a
n
d
Re
c
o
g
n
it
io
n
i
n
V
isu
a
l
S
u
rv
e
il
lan
c
e
,
”
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
C
o
mp
u
t
e
r E
n
g
i
n
e
e
rin
g
(
IJ
ECE
)
;
V
o
l.
2
,
N
o
.
3
,
p
p
.
3
9
5
~
4
0
4
,
Ju
n
e
2
0
1
2
.
[3
3
]
S
.
W
u
,
B.
E.
M
o
o
re
,
a
n
d
M
.
S
h
a
h
,
“
Ch
a
o
ti
c
in
v
a
rian
ts
o
f
lag
ra
n
g
i
a
n
p
a
rti
c
le
traje
c
to
ries
f
o
r
a
n
o
m
a
ly
d
e
tec
ti
o
n
in
c
ro
w
d
e
d
sc
e
n
e
s,”
in
IEE
E
C
o
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
Vi
sio
n
a
n
d
Pa
t
ter
n
Rec
o
g
n
it
io
n
,
C
VPR,
p
p
.
2
0
5
4
–
2
0
6
0
,
2
0
1
0
.
[3
4
]
R.
M
e
h
r
a
n
,
A
.
Oy
a
m
a
,
a
n
d
M
.
S
h
a
h
,
“
A
b
n
o
rm
a
l
c
ro
w
d
b
e
h
a
v
io
r
d
e
tec
ti
o
n
u
sin
g
so
c
ial
f
o
rc
e
m
o
d
e
l
,
”
i
n
IEE
E
Co
n
fer
e
n
c
e
o
n
Co
m
p
u
ter
Vi
si
o
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
C
VP
R,
p
p
.
9
3
5
–
9
4
2
,
2
0
0
9
.
[3
5
]
Y.
Co
n
g
,
J.
Y
u
a
n
,
a
n
d
J.
L
iu
,
“
S
p
a
rse
re
c
o
n
stru
c
ti
o
n
c
o
st
f
o
r
a
b
n
o
rm
a
l
e
v
e
n
t
d
e
tec
ti
o
n
.
”
i
n
C
o
mp
u
ter
Vi
si
o
n
a
n
d
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
C
VP
R
,
2
0
1
1
,
p
p
.
3
4
4
9
–
3
4
5
6
.
[3
6
]
V
.
S
a
li
g
ra
m
a
a
n
d
Z.
Ch
e
n
,
“
Vi
d
e
o
a
n
o
m
a
ly d
e
tec
ti
o
n
b
a
se
d
o
n
l
o
c
a
l
st
a
ti
stica
l
a
g
g
re
g
a
tes
,
”
in
IEE
E
Co
n
fer
e
n
c
e
o
n
Co
mp
u
ter
V
isio
n
a
n
d
Pa
t
ter
n
Rec
o
g
n
it
io
n
,
C
VPR,
p
p
.
2
1
1
2
–
2
1
1
9
,
2
0
1
2
.
[3
7
]
W
.
L
i,
V
.
M
a
h
a
d
e
v
a
n
,
a
n
d
N.
V
a
sc
o
n
c
e
lo
s,
“
A
n
o
m
a
l
y
d
e
tec
ti
o
n
a
n
d
lo
c
a
li
z
a
ti
o
n
i
n
c
ro
w
d
e
d
sc
e
n
e
s
,
”
IEE
E
T
ra
n
s.
Pa
tt
e
rn
A
n
a
lys
is a
n
d
M
a
c
h
i
n
e
In
t
e
ll
ig
e
n
c
e
,
v
o
l.
3
6
,
n
o
.
1
,
p
p
.
1
8
–
3
2
,
2
0
1
4
.
[3
8
]
V
.
Ka
lt
sa
,
A
.
Brias
so
u
li
,
I.
Ko
m
p
a
tsiaris,
L
.
J.
Ha
d
ji
leo
n
ti
a
d
is
,
a
n
d
M
.
G
.
S
tri
n
tzis,
“
S
w
a
r
m
in
telli
g
e
n
c
e
f
o
r
d
e
tec
ti
n
g
in
tere
stin
g
e
v
e
n
ts
in
c
ro
w
d
e
d
e
n
v
iro
n
m
e
n
ts,”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Ima
g
e
Pro
c
e
ss
in
g
,
v
o
l
.
2
4
,
n
o
.
7
,
p
p
.
2
1
5
3
–
2
1
6
6
,
2
0
1
5
.
[3
9
]
H.
F
ra
d
i;
B.
L
u
v
iso
n
;
Q.
C
.
P
h
a
m
,
"
Cro
w
d
Be
h
a
v
io
r
A
n
a
ly
sis
Us
in
g
L
o
c
a
l
M
id
-
L
e
v
e
l
V
isu
a
l
De
sc
ri
p
to
rs,"
i
n
I
EE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Circ
u
it
s
a
n
d
S
y
ste
ms
fo
r
Vi
d
e
o
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
P
P
,
n
o
.
9
9
,
p
p
.
1
-
1
d
o
i:
1
0
.
1
1
0
9
/T
CS
V
T
.
2
0
1
6
.
2
6
1
5
4
4
3
,
2
0
1
6
.
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
C
r
o
w
d
r
ec
o
g
n
itio
n
s
ystem
b
a
s
ed
o
n
o
p
tica
l flo
w
a
lo
n
g
w
ith
S
V
M c
la
s
s
ifier
(
S
h
r
ee
d
a
r
s
h
a
n
.
K
)
2459
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
S
h
r
e
e
d
a
r
s
h
a
n
.
K
is
A
ss
istan
t
P
r
o
f
e
ss
o
r
in
E&C
d
e
p
a
rt
m
e
n
t
a
t
M
.
S
.
Ra
m
a
iah
In
sti
tu
te
o
f
Tec
h
n
o
lo
g
y
Ba
n
g
a
lo
re
.
He
G
ra
d
u
a
ted
w
it
h
B.
E.
d
e
g
re
e
in
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
E
n
g
in
e
e
rin
g
.
H
e
d
i
d
M
.
T
e
c
h
w
it
h
sp
e
c
ializa
ti
o
n
in
Di
g
it
a
l
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
s
f
ro
m
M
a
n
ip
a
l
Un
iv
e
rsity
.
He
is
c
u
rre
n
tl
y
w
o
rk
in
g
to
wa
rd
s
th
e
P
h
.
D.
d
e
g
re
e
.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
im
a
g
e
p
ro
c
e
ss
in
g
a
n
d
c
r
y
p
to
g
ra
p
h
y
.
S
.
S
e
th
u
S
e
lv
i
is w
o
rk
in
g
a
s th
e
P
ro
f
e
ss
o
r
&
He
a
d
o
f
th
e
El
e
c
tro
n
ics
a
n
d
Co
m
m
u
n
ica
ti
o
n
De
p
a
rtm
e
n
t
o
f
M
S
RIT
.
S
h
e
g
ra
d
u
a
ted
w
it
h
B.
E
in
El
e
c
tro
n
ics
a
n
d
C
o
m
m
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
f
ro
m
M
a
d
u
ra
i
Ka
m
a
ra
j
Un
iv
e
rsit
y
,
M
E
f
ro
m
A
n
n
a
Un
iv
e
rsity
,
a
n
d
P
h
.
D
in
S
ig
n
a
l
P
ro
c
e
ss
in
g
f
ro
m
IIS
C,
Ba
n
g
a
lo
re
.
He
r
su
b
jec
ts
o
f
in
tere
sts a
re
sig
n
a
l
p
ro
c
e
ss
in
g
,
im
a
g
e
p
ro
c
e
ss
in
g
a
n
d
d
ig
it
a
l
c
o
m
m
u
n
ica
ti
o
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.