I
nte
rna
t
io
na
l
J
o
urna
l
of
I
nfo
r
m
a
t
ics
a
nd
Co
mm
u
n
ica
t
io
n
T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
8
,
No
.
1
,
A
p
r
il
201
9
,
p
p
.
1
~
12
I
SS
N:
2252
-
8776
,
DOI
:
1
0
.
1
1
5
9
1
/i
j
ict.
v
8
i1
.
p
p
1
-
12
1
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
JI
C
T
A
no
v
el
s
k
et
ch
-
ba
sed
3D
m
o
del
retr
iev
a
l
a
ppro
a
ch
ba
sed
o
n
s
k
eleton
J
ing
Z
ha
ng
,
B
a
o
s
heng
K
a
n
g
,
Bo
J
ia
ng
,
Di
Z
ha
ng
No
rth
w
e
st
Un
iv
e
rsit
y
/De
p
a
rt
m
e
n
t
of
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
Ch
in
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Oct
9,
20
18
R
ev
i
s
ed
Dec
20,
2
0
1
8
A
cc
ep
ted
J
an
11,
2
0
1
9
S
in
c
e
th
e
sk
e
leto
n
re
p
re
se
n
ts
th
e
to
p
o
l
o
g
y
stru
c
tu
re
of
th
e
q
u
e
ry
sk
e
tch
a
n
d
2D
v
iew
s
of
3D
m
o
d
e
l,
t
h
is
p
a
p
e
r
p
ro
p
o
se
s
a
n
o
v
e
l
sk
e
tch
-
b
a
se
d
3D
m
o
d
e
l
re
tri
e
v
a
l
a
lg
o
rit
h
m
w
h
ich
u
ti
li
z
e
s
sk
e
leto
n
c
h
a
ra
c
teristics
as
th
e
f
e
a
tu
re
s
to
d
e
sc
rib
e
th
e
o
b
jec
t
sh
a
p
e
.
F
irstl
y
,
we
p
ro
p
o
se
a
d
v
a
n
c
e
d
sk
e
leto
n
stre
n
g
t
h
map
(
A
S
S
M
)
a
lg
o
rit
h
m
to
c
re
a
te
th
e
sk
e
l
e
to
n
w
h
ich
c
o
m
p
u
tes
th
e
sk
e
leto
n
stre
n
g
th
m
a
p
by
iso
tro
p
ic
d
if
f
u
sio
n
on
t
h
e
g
ra
d
ien
t
v
e
c
to
r
f
ield
,
se
lec
ts
c
rit
ica
l
p
o
in
ts
f
ro
m
th
e
s
k
e
let
o
n
stre
n
g
th
m
a
p
a
n
d
c
o
n
n
e
c
ts
th
e
m
by
Kru
sk
a
l'
s
a
l
g
o
rit
h
m
.
T
h
e
n
,
we
p
ro
p
o
se
h
isto
g
ra
m
f
e
a
tu
re
c
o
m
p
a
riso
n
a
lg
o
rit
h
m
w
h
ich
a
d
o
p
ts
th
e
ra
d
ii
of
th
e
d
isk
s
at
s
k
e
leto
n
p
o
in
ts
a
n
d
th
e
len
g
th
s
of
sk
e
leto
n
b
ra
n
c
h
e
s
to
e
x
trac
t
th
e
h
isto
g
ra
m
f
e
a
tu
re
,
a
nd
c
o
m
p
a
re
th
e
sim
il
a
rit
y
b
e
t
w
e
e
n
t
w
o
sk
e
le
to
n
s
u
sin
g
th
e
h
isto
g
ra
m
f
e
a
tu
re
m
a
tri
x
of
sk
e
leto
n
e
n
d
p
o
in
ts.
Ex
p
e
rim
e
n
t
re
su
lt
s
d
e
m
o
n
stra
te
th
a
t
our
a
p
p
ro
a
c
h
w
h
ich
c
o
m
b
in
e
s
th
e
se
t
w
o
a
lg
o
rit
h
m
s
sig
n
if
ica
n
tl
y
o
u
tp
e
rf
o
r
m
s
se
v
e
r
a
l
lea
d
in
g
sk
e
tch
-
b
a
se
d
re
tri
e
v
a
l
a
p
p
ro
a
c
h
e
s
.
K
ey
w
o
r
d
s
:
A
d
v
an
ce
d
s
k
eleto
n
s
tr
en
g
t
h
m
ap
His
to
g
r
a
m
f
ea
tu
r
e
co
m
p
ar
is
o
n
Sk
eleto
n
f
ea
t
u
r
es
S
k
etc
h
-
b
ased
3D
M
o
d
el
R
etr
iev
al
Co
p
y
rig
h
t
©
2019
In
stit
u
te
of
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
.
All
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
J
in
g
Z
h
a
n
g,
Dep
ar
t
m
en
t
of
I
n
f
o
r
m
atio
n
Sc
ien
ce
an
d
T
ec
h
n
o
lo
g
y
,
No
r
th
w
est
U
n
i
v
er
s
it
y
,
C
h
a
n
g
An
Q
u
.
,
Xi
'
an
,
C
h
in
a
.
E
m
ail:
m
y
w
an
g
w
a
n
g
1
2
3
@
1
6
3
.
co
m
1.
I
NT
RO
D
UCT
I
O
N
W
ith
t
h
e
r
ap
id
in
cr
ea
s
e
i
n
t
h
e
n
u
m
b
er
o
f
a
v
ailab
le
3
D
m
o
d
els,
ac
cu
r
atel
y
a
n
d
ef
f
ec
ti
v
el
y
s
ea
r
ch
3
D
m
o
d
el
i
s
i
m
p
o
r
tan
t
f
o
r
m
an
y
a
p
p
licatio
n
s
s
u
ch
as
i
n
d
u
s
tr
ial
d
esig
n
,
en
g
in
ee
r
i
n
g
,
an
d
m
a
n
u
f
ac
t
u
r
in
g
ar
ea
.
T
h
e
m
o
s
t
p
o
p
u
lar
w
a
y
f
o
r
r
etr
iev
i
n
g
3
D
m
o
d
els
is
e
x
a
m
p
le
-
b
as
ed
p
ar
ad
ig
m
,
w
h
er
e
th
e
u
s
er
p
r
o
v
id
es
an
e
x
is
t
in
g
3
D
m
o
d
el
as
q
u
er
y
i
n
p
u
t
an
d
th
e
r
etr
iev
a
l
s
y
s
te
m
c
a
n
r
e
tu
r
n
s
i
m
ilar
3
D
m
o
d
els
f
r
o
m
th
e
d
atab
ase
[
1
]
.
Ho
w
e
v
er
,
it
i
s
d
if
f
ic
u
lt
f
o
r
a
u
s
er
to
h
a
v
e
a
n
ap
p
r
o
p
r
iate
ex
am
p
le
3
D
m
o
d
el
a
t
h
a
n
d
.
An
al
ter
n
ati
v
e
w
a
y
is
to
u
s
e
2
D
s
k
etch
as
a
q
u
er
y
w
h
e
r
e
u
s
er
s
ca
n
d
escr
ib
e
a
tar
g
et
3
D
m
o
d
el
b
y
q
u
ic
k
l
y
d
r
a
w
i
n
g
it.
B
u
t
a
2
D
s
k
etc
h
is
m
er
el
y
a
co
ar
s
e
a
n
d
s
i
m
p
le
r
ep
r
esen
tatio
n
w
h
ic
h
o
n
l
y
co
n
tain
s
p
ar
tial i
n
f
o
r
m
atio
n
o
f
a
n
o
r
ig
in
al
3
D
m
o
d
el.
Hen
ce
,
it is
m
o
r
e
ch
alle
n
g
in
g
t
o
r
ea
lize
a
s
k
etch
-
b
ased
r
etr
iev
al
th
a
n
a
n
ex
a
m
p
le
-
b
ased
r
etr
iev
al
[
2
]
.
T
h
e
s
k
eleto
n
is
w
id
el
y
r
ec
o
g
n
iz
ed
as
o
n
e
o
f
th
e
m
o
s
t
i
m
p
o
r
tan
t
s
h
ap
e
d
escr
ip
to
r
s
in
s
k
etch
-
b
ased
3
D
m
o
d
el
r
etr
iev
al
p
r
o
ce
s
s
.
T
h
e
s
k
eleto
n
e
x
tr
ac
tio
n
a
lg
o
r
ith
m
s
g
e
n
er
all
y
ar
e
cla
s
s
i
f
ied
in
to
f
iv
e
t
y
p
es:
t
h
e
th
i
n
n
i
n
g
al
g
o
r
ith
m
[
3
]
,
t
h
e
d
is
cr
ete
d
o
m
ain
al
g
o
r
ith
m
b
ased
o
n
t
h
e
Vo
r
o
n
o
i
d
iag
r
a
m
s
[
4
]
,
th
e
a
n
al
y
tic
co
m
p
u
tatio
n
o
f
th
e
o
b
j
ec
t
co
n
to
u
r
alg
o
r
it
h
m
[
5
]
,
th
e
al
g
o
r
ith
m
b
ased
o
n
m
at
h
e
m
atica
l
m
o
r
p
h
o
lo
g
y
[
6
]
,
an
d
th
e
al
g
o
r
ith
m
b
ased
o
n
d
i
s
tan
ce
tr
an
s
f
o
r
m
[
7
]
.
B
er
tr
an
d
G
et
al
[
8
]
f
o
u
n
d
t
h
at
t
h
e
t
h
i
n
n
in
g
al
g
o
r
ith
m
is
q
u
it
e
s
en
s
iti
v
e
to
n
o
i
s
e
a
n
d
ca
n
ac
h
iev
e
lar
g
e
a
m
o
u
n
t
o
f
ca
lc
u
l
atio
n
,
w
h
ich
is
e
x
p
ec
ted
to
eli
m
i
n
ate
lo
ca
l
n
o
is
e
w
it
h
o
u
t
i
n
tr
o
d
u
cin
g
d
is
to
r
tio
n
s
an
d
r
etai
n
s
i
g
n
i
f
ican
t
f
ea
t
u
r
es
o
f
t
h
e
p
atter
n
.
G
iese
n
J
et
al
[
9
]
r
esear
ch
ed
o
n
th
e
d
is
cr
ete
d
o
m
ai
n
alg
o
r
it
h
m
b
ased
o
n
th
e
Vo
r
o
n
o
i
d
iag
r
a
m
s
w
h
ich
ca
lc
u
lates
t
h
e
Vo
r
o
n
o
i
d
iag
r
a
m
s
v
ia
a
s
u
f
f
icie
n
t la
r
g
e
s
et
o
f
p
o
in
t
s
s
a
m
p
led
f
r
o
m
its
b
o
u
n
d
ar
y
.
I
f
t
h
e
a
m
o
u
n
t o
f
s
a
m
p
led
p
o
in
ts
is
ex
tr
e
m
e
l
y
lar
g
e,
it
w
o
u
ld
lead
to
th
e
n
u
m
er
ical
i
n
s
tab
ili
t
y
i
n
th
e
co
m
p
u
tatio
n
o
f
Vo
r
o
n
o
i
d
iag
r
am
s
.
Kr
in
id
i
s
S
et
al
[
1
0
]
f
o
u
n
d
th
at
t
h
e
a
n
al
y
tic
co
m
p
u
tatio
n
o
f
t
h
e
o
b
j
ec
t
co
n
to
u
r
al
g
o
r
ith
m
w
o
u
ld
lead
to
t
h
e
al
g
o
r
ith
m
’
s
in
ac
c
u
r
ac
y
d
u
e
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8776
IJ
-
I
C
T
Vo
l.
8
,
No
.
1
,
A
p
r
il
2
0
1
9
:
1
–
12
2
th
e
p
o
l
y
g
o
n
al
ap
p
r
o
x
i
m
atio
n
is
m
o
s
tl
y
n
o
t
s
u
f
f
icie
n
t
f
o
r
b
io
lo
g
ical
s
h
ap
e
d
escr
ip
tio
n
.
Mitr
a
A
et
al
[
1
1
]
f
o
u
n
d
th
a
t
t
h
e
al
g
o
r
ith
m
b
ase
d
o
n
m
at
h
e
m
atica
l
m
o
r
p
h
o
lo
g
y
ca
n
lo
ca
lize
t
h
e
ac
c
u
r
ate
s
k
eleto
n
,
b
u
t
m
a
y
n
o
t
g
u
ar
a
n
tee
t
h
e
e
x
ac
t
co
n
n
ec
ti
v
it
y
o
f
th
e
s
k
eleto
n
.
Ya
n
T
Q
et
al
[
1
2
]
f
o
u
n
d
th
a
t
t
h
e
alg
o
r
ith
m
b
ased
o
n
d
is
tan
ce
tr
a
n
s
f
o
r
m
co
m
p
u
te
t
h
e
s
k
eleto
n
b
y
d
etec
ti
n
g
r
id
g
es
o
n
th
e
d
i
s
ta
n
ce
tr
an
s
f
o
r
m
s
u
r
f
ac
e,
w
h
ic
h
en
s
u
r
e
s
th
e
ac
c
u
r
ate
lo
ca
lizatio
n
o
f
s
k
eleto
n
p
o
i
n
ts
.
Ho
w
e
v
er
,
th
e
alg
o
r
ith
m
b
ased
o
n
d
i
s
tan
ce
t
r
an
s
f
o
r
m
is
lack
o
f
co
n
n
ec
ti
v
it
y
a
n
d
co
m
p
leten
e
s
s
,
w
h
ich
m
ea
n
s
,
th
e
ex
tr
ac
ted
b
r
an
ch
es
m
a
y
b
e
d
is
co
n
n
ec
te
d
an
d
n
o
t
b
e
ab
le
t
o
r
ep
r
e
s
en
t a
ll th
e
s
ig
n
i
f
ica
n
t v
i
s
u
al
p
ar
ts
.
Ho
w
to
co
m
p
ar
e
th
e
s
k
e
leto
n
s
i
m
ilar
it
y
b
et
w
ee
n
t
h
e
q
u
er
y
s
k
etc
h
a
n
d
2
D
v
ie
w
s
o
f
3
D
m
o
d
el
i
s
cr
u
cial.
T
h
er
e
ar
e
t
h
r
ee
t
y
p
es
o
f
a
lg
o
r
it
h
m
s
to
co
m
p
ar
e
t
h
e
s
k
eleto
n
s
i
m
ilar
it
y
,
r
e
s
p
ec
ti
v
e
l
y
tr
ee
m
atc
h
i
n
g
alg
o
r
ith
m
[
1
3
]
,
p
ath
s
i
m
ilar
it
y
alg
o
r
it
h
m
[
1
4
]
an
d
g
r
ap
h
r
ep
r
esen
tatio
n
al
g
o
r
ith
m
[
1
5
]
.
Dr
ag
an
F
F
et
al
[
1
6
]
f
o
u
n
d
th
at
th
e
tr
ee
d
escr
ip
to
r
r
ep
r
esen
ts
t
h
e
to
p
o
lo
g
ical
f
ea
t
u
r
es
o
f
t
h
e
s
k
eleto
n
,
b
u
t
t
h
e
m
ax
i
m
a
l
is
o
m
o
r
p
h
ic
s
u
b
tr
ee
s
ar
e
o
b
tain
ed
b
y
s
ea
r
ch
in
g
f
o
r
th
e
lo
n
g
est
m
atc
h
i
n
g
s
u
b
s
tr
i
n
g
s
w
h
ic
h
w
o
u
ld
l
ea
d
to
h
ig
h
er
ti
m
e
co
m
p
le
x
it
y
.
B
ai
X
et
al
[
1
4
]
m
atc
h
ed
th
e
s
k
e
leto
n
g
r
ap
h
s
b
y
co
m
p
ar
in
g
t
h
e
g
eo
d
esic
p
a
th
s
a
m
o
n
g
s
k
e
leto
n
en
d
p
o
in
ts
.
Ho
w
e
v
er
,
th
eir
ap
p
r
o
ac
h
is
o
n
ly
m
o
t
iv
ated
b
y
t
h
e
v
i
s
u
a
ll
y
s
i
m
i
lar
s
k
elet
o
n
g
r
ap
h
s
,
w
it
h
o
u
t
co
n
s
id
er
in
g
th
e
to
p
o
lo
g
ical
g
r
ap
h
s
tr
u
ct
u
r
e.
Yo
u
s
s
ef
R
et
al
[
1
7
]
f
o
u
n
d
th
at
th
e
g
r
ap
h
r
ep
r
esen
tatio
n
alg
o
r
ith
m
is
w
id
el
y
i
n
v
e
s
ti
g
a
ted
r
eg
ar
d
in
g
m
atc
h
i
n
g
is
s
u
e
s
,
d
u
e
to
th
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
s
k
eleto
n
b
r
an
ch
es
w
it
h
g
r
ap
h
ed
g
es
a
n
d
n
o
d
es
is
n
a
tu
r
al
a
n
d
i
n
t
u
it
iv
e.
Ho
w
e
v
er
,
t
h
e
s
o
lu
tio
n
s
i
n
t
h
e
liter
atu
r
e
ar
e
b
ased
o
n
s
h
o
ck
g
r
ap
h
s
[
1
8
]
o
r
attr
ib
u
ted
r
elatio
n
al
g
r
ap
h
s
[
1
9
]
w
h
ic
h
ar
e
o
n
l
y
u
s
ed
i
n
ce
r
tain
ap
p
licatio
n
d
o
m
ai
n
s
a
n
d
u
s
u
all
y
w
it
h
h
ig
h
er
ti
m
e
co
m
p
le
x
it
y
.
T
o
tack
le
th
ese
p
r
o
b
le
m
s
,
t
h
i
s
p
ap
er
p
r
o
p
o
s
es
a
n
o
v
el
s
k
e
tch
-
b
ased
3
D
m
o
d
el
r
etr
iev
al
ap
p
r
o
ac
h
w
h
ic
h
co
m
b
i
n
es
t
h
e
ad
v
a
n
ce
d
s
k
eleto
n
s
tr
e
n
g
th
m
ap
(
A
S
S
M)
alg
o
r
ith
m
w
it
h
h
is
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
alg
o
r
ith
m
to
cr
ea
te
t
h
e
s
k
eleto
n
an
d
co
m
p
ar
e
t
h
e
s
i
m
i
lar
it
y
.
Ou
r
A
S
SM
alg
o
r
it
h
m
o
r
ig
i
n
s
f
r
o
m
t
h
e
al
g
o
r
ith
m
b
ased
o
n
d
is
tan
ce
tr
an
s
f
o
r
m
,
w
h
ic
h
en
s
u
r
es
t
h
e
ac
cu
r
ate
lo
ca
lizatio
n
o
f
s
k
eleto
n
p
o
in
t
s
.
Fu
r
t
h
er
,
o
u
r
A
SS
M
alg
o
r
ith
m
s
elec
t
s
cr
itical
p
o
in
ts
f
r
o
m
t
h
e
s
k
eleto
n
s
tr
en
g
th
m
ap
an
d
co
n
n
ec
t
s
th
e
m
b
y
K
r
u
s
k
al
'
s
al
g
o
r
ith
m
[
2
0
]
,
w
h
ic
h
s
o
l
v
es
t
h
e
co
n
n
ec
tiv
it
y
an
d
co
m
p
lete
n
es
s
p
r
o
b
l
e
m
o
f
p
r
ev
io
u
s
alg
o
r
it
h
m
.
T
h
e
h
is
to
g
r
a
m
f
ea
tu
r
e
co
m
p
ar
is
o
n
alg
o
r
it
h
m
o
r
ig
i
n
s
f
r
o
m
th
e
g
r
ap
h
r
ep
r
esen
tatio
n
alg
o
r
ith
m
,
w
h
ich
ex
p
r
ess
es
t
h
e
s
k
eleto
n
b
r
an
ch
es
n
at
u
r
all
y
a
n
d
in
tu
i
ti
v
el
y
[
2
1
]
.
Ou
r
h
is
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
al
g
o
r
ith
m
ad
o
p
ts
th
e
r
ad
ii
o
f
t
h
e
d
is
k
s
at
s
k
ele
to
n
p
o
in
ts
a
n
d
th
e
len
g
t
h
s
o
f
s
k
eleto
n
b
r
an
c
h
es
to
ex
tr
ac
t
th
e
h
i
s
to
g
r
a
m
f
ea
tu
r
e.
T
h
e
r
ad
ii
o
f
th
e
d
is
k
s
an
d
len
g
t
h
s
o
f
s
k
e
leto
n
b
r
an
c
h
es
ar
e
i
n
v
ar
ia
n
t
u
n
d
er
th
e
e
n
v
ir
o
n
m
en
t
o
f
n
o
n
-
r
i
g
id
tr
an
s
f
o
r
m
atio
n
s
w
h
ic
h
w
o
u
ld
h
elp
t
o
g
et
h
i
g
h
-
p
r
ec
is
io
n
r
etr
iev
al
r
e
s
u
l
t.
A
d
d
itio
n
all
y
,
w
e
co
m
p
ar
e
t
h
e
s
i
m
i
lar
it
y
b
et
w
ee
n
t
w
o
s
k
eleto
n
s
u
s
in
g
th
e
h
i
s
to
g
r
a
m
f
ea
t
u
r
e
m
a
tr
ix
o
f
s
k
e
leto
n
e
n
d
p
o
in
ts
,
w
it
h
r
elat
iv
el
y
less
q
u
a
n
tit
y
t
h
a
n
s
k
eleto
n
p
o
in
ts
,
w
h
ic
h
lead
s
to
a
lo
w
er
co
m
p
u
tatio
n
al
co
m
p
le
x
i
t
y
.
T
o
ev
alu
ate
o
u
r
ap
p
r
o
ac
h
,
w
e
t
est
o
u
r
ap
p
r
o
ac
h
o
n
th
e
p
u
b
lic
s
ta
n
d
ar
d
d
ataset
an
d
also
co
m
p
ar
ed
w
it
h
o
t
h
er
lead
in
g
3
D
m
o
d
el
r
etr
ie
v
al
ap
p
r
o
ac
h
es.
T
h
e
ex
p
er
i
m
e
n
ts
d
e
m
o
n
s
tr
ate
th
at
o
u
r
ap
p
r
o
ac
h
is
s
ig
n
i
f
ica
n
tl
y
b
etter
th
an
s
e
v
er
al
lead
i
n
g
s
k
etch
-
b
ased
r
e
tr
iev
a
l
ap
p
r
o
ac
h
es.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
e
f
r
a
m
e
w
o
r
k
of
s
k
etch
-
b
as
ed
3D
m
o
d
el
r
etr
iev
al
b
ased
on
ad
v
a
n
ce
d
s
k
eleto
n
s
tr
e
n
g
t
h
m
ap
an
d
h
is
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
alg
o
r
ith
m
is
p
r
o
p
o
s
ed
as
s
h
o
wn
in
Fig
u
r
e
1.
2
D
Q
u
e
r
y
S
k
e
t
c
h
I
n
p
u
t
S
t
a
r
t
3
D
M
o
d
e
l
S
e
t
2
D
P
r
o
j
e
c
t
i
o
n
V
i
e
w
o
f
3
D
M
o
d
e
l
H
i
s
t
o
g
r
a
m
F
e
a
t
u
r
e
E
x
t
r
a
c
t
i
o
n
S
k
e
l
e
t
o
n
D
i
s
t
a
n
c
e
C
o
m
p
u
t
a
t
i
o
n
2
D
P
r
o
j
e
c
t
i
o
n
V
i
e
w
F
e
a
t
u
r
e
S
e
t
R
e
t
r
i
e
v
a
l
R
e
s
u
l
t
s
E
n
d
A
d
v
a
n
c
e
d
S
k
e
l
e
t
o
n
S
t
r
e
n
g
t
h
M
a
p
H
i
s
t
o
g
r
a
m
F
e
a
t
u
r
e
E
x
t
r
a
c
t
i
o
n
2
D
S
k
e
t
c
h
F
e
a
t
u
r
e
S
e
t
A
d
v
a
n
c
e
d
S
k
e
l
e
t
o
n
S
t
r
e
n
g
t
h
M
a
p
Fig
u
r
e
1.
T
h
e
f
r
am
e
w
o
r
k
of
o
u
r
s
k
e
tch
-
b
ased
3D
m
o
d
el
r
etr
iev
al
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
A
n
o
ve
l
s
ke
tch
-
b
a
s
ed
3D
mo
d
el
r
etri
ev
a
l
a
p
p
r
o
a
ch
b
a
s
ed
on
s
ke
leto
n
(
Jin
g
Zh
a
n
g
)
3
2
.
1
.
P
re
-
pro
ce
s
s
ing
I
n
th
e
d
atab
ase
ea
c
h
3
D
m
o
d
e
l
h
as
a
n
ar
b
itra
r
y
p
o
s
itio
n
,
o
r
i
en
tatio
n
a
n
d
s
ca
le
i
n
t
h
e
s
p
ati
al
s
p
ac
e,
it
is
n
ec
es
s
ar
y
to
n
o
r
m
alize
ea
ch
3
D
m
o
d
el
b
ef
o
r
e
p
r
o
j
ec
t
th
em
in
to
2
D
v
ie
w
s
.
Af
ter
3
D
m
o
d
e
ls
in
th
e
d
atab
ase
h
av
e
b
ee
n
n
o
r
m
alize
d
,
w
e
co
m
p
ar
e
t
h
e
q
u
er
y
s
k
e
tch
w
it
h
6
0
p
r
o
j
ec
tio
n
v
ie
w
s
o
f
ea
ch
3
D
m
o
d
el
u
s
in
g
li
g
h
t
f
ield
d
escr
ip
to
r
s
[
2
2
]
.
I
n
ad
d
itio
n
,
w
e
u
t
ilize
2
D
s
k
etc
h
-
3
D
m
o
d
el
ali
g
n
m
e
n
t
al
g
o
r
it
h
m
to
s
elec
t
s
o
m
e
ca
n
d
id
ate
v
ie
w
s
f
r
o
m
t
h
e
2
D
p
r
o
j
ec
tio
n
v
ie
w
s
f
o
r
ef
f
icien
tl
y
co
m
p
ar
in
g
t
h
e
s
i
m
ilar
it
y
.
Ou
r
lig
h
t
f
ield
s
ar
e
d
is
tr
ib
u
te
d
u
n
i
f
o
r
m
l
y
a
n
d
p
o
s
itio
n
ed
on
v
er
tice
s
of
a
r
eg
u
lar
d
o
d
ec
ah
ed
r
o
n
.
Fo
r
r
eg
u
lar
d
o
d
ec
ah
ed
r
o
n
,
each
of
th
e
20
v
er
tice
s
is
co
n
n
ec
ted
by
3
ed
g
es,
w
h
ich
r
e
s
u
l
ts
in
60
d
if
f
er
e
nt
r
o
tatio
n
s
(
3
edg
e
20
s
ve
r
t
i
ce
s
)
f
o
r
o
u
r
p
r
o
j
ec
tio
n
s
,
so
th
at
we
n
ee
d
r
o
tate
th
e
ca
m
er
a
s
y
s
te
m
60
ti
m
e
s
.
W
h
en
th
e
ca
m
er
as
s
w
itch
o
n
to
d
if
f
er
en
t
v
er
tices
[
2
3
]
we
m
ea
s
u
r
e
th
e
s
i
m
ilar
it
y
b
et
w
ee
n
3D
m
o
d
el
a
nd
s
k
e
tch
.
Her
e
is
a
t
y
p
ical
ex
a
m
p
le
to
ex
p
lai
n
o
u
r
ap
p
r
o
ac
h
.
As
s
h
o
w
n
in
Fi
g
u
r
e
2,
in
t
h
e
le
f
t
is
a
p
ig
3D
m
o
d
el,
in
t
h
e
r
ig
h
t
d
is
p
la
y
60
p
r
o
j
ec
tio
n
v
ie
w
s
wh
ich
ar
e
r
en
d
er
ed
f
r
o
m
v
er
tice
s
of
a
d
o
d
ec
ah
ed
r
o
n
f
o
r
th
e
p
ig
3D
m
o
d
el.
Fig
u
r
e
2
.
T
h
e
lig
h
t
fi
eld
d
escr
ip
to
r
s
of
3D
m
o
d
el
p
r
o
j
ec
tio
n
To
en
h
an
ce
t
h
e
r
etr
iev
a
l
ac
cu
r
ac
y
a
n
d
p
er
f
o
r
m
an
ce
,
we
ad
o
p
t
2D
s
k
etch
-
3D
m
o
d
el
alig
n
m
e
n
t
alg
o
r
ith
m
[
2
4
]
to
ch
o
o
s
e
th
e
ca
n
d
id
ate
v
ie
w
s
in
th
e
s
k
etc
h
-
b
ased
3D
m
o
d
el
r
etr
iev
al.
We
ch
o
o
s
e
ca
n
d
id
ate
v
ie
w
s
[
2
5
]
by
k
ee
p
in
g
a
ce
r
tain
p
er
ce
n
ta
g
e
T
w
it
h
to
p
s
i
m
ilar
ities
b
et
w
ee
n
t
h
e
s
k
et
ch
an
d
all
th
e
2D
p
r
o
j
ec
tio
n
v
ie
w
s
,
e.
g
.
20%
T
m
ea
n
s
t
h
at
th
e
n
u
m
b
er
of
o
u
r
ca
n
d
id
a
te
v
ie
w
s
is
6
0
*
2
0
%
=
1
2
.
2
.
2
.
Adv
a
nced
s
k
elet
o
n
s
t
re
ng
t
h
m
a
p
(
A
SS
M
)
T
h
e
alg
o
r
ith
m
of
s
k
eleto
n
e
x
tr
ac
tio
n
by
A
S
SM
is
d
escr
ib
ed
as
f
o
llo
w
s
:
1.
Step
1.
E
x
tr
ac
tio
n
of
e
x
ter
n
al
b
o
u
n
d
ar
y
.
Fo
r
each
q
u
er
y
s
k
et
ch
an
d
2D
v
ie
w
s
of
3D
m
o
d
el
s
,
th
e
e
x
ter
n
a
l
b
o
u
n
d
ar
y
is
e
x
tr
ac
ted
f
ir
s
t
l
y
,
w
h
ic
h
p
r
o
v
id
es
m
u
c
h
of
th
e
i
m
ag
e
v
i
s
u
al
i
n
f
o
r
m
atio
n
f
o
r
th
e
SSM
v
al
u
e.
2.
Step
2
.
C
o
m
p
u
tat
io
n
o
f
SS
M
v
alu
e
[
2
6
]
.
W
e
p
er
f
o
r
m
d
is
tan
ce
tr
an
s
f
o
r
m
o
n
b
o
u
n
d
ar
y
a
n
d
th
en
co
m
p
u
t
e
th
e
SS
M
v
al
u
e
b
y
is
o
tr
o
p
ic
d
if
f
u
s
io
n
o
n
th
e
g
r
ad
ien
t
v
ec
to
r
f
ield
.
3.
Step
3.
R
e
f
in
ed
SSM
v
al
u
e.
We
ad
o
p
t
th
e
n
o
n
-
m
a
x
i
m
al
s
u
p
p
r
ess
io
n
al
g
o
r
ith
m
[
2
7
]
f
o
r
t
h
e
r
ef
in
ed
SSM
v
alu
e.
4.
Step
4.
Selectio
n
of
cr
itical
p
o
in
ts
.
We
s
elec
t
t
h
e
cr
itical
p
o
in
ts
f
r
o
m
t
h
e
r
ef
i
n
ed
SSM
v
al
u
e.
5.
Step
5.
Sk
ele
to
n
tr
ac
e.
T
h
e
f
i
n
al
s
k
ele
to
n
is
o
b
tain
ed
by
c
o
n
n
ec
ti
n
g
t
h
e
cr
itical
p
o
in
t
s
.
We
p
r
o
p
o
s
e
a
co
n
n
ec
ti
n
g
cr
itical
p
o
in
ts
’
m
et
h
o
d
w
h
ic
h
u
s
es
th
e
Kr
u
s
k
al
's
al
g
o
r
ith
m
to
d
ec
id
e
t
h
e
o
r
d
er
of
th
e
co
n
n
ec
ti
n
g
p
ath
.
2
.
2
.
1.
Co
m
pu
t
a
t
io
n
of
SSM
v
a
lue
We
d
ef
in
e
a
f
u
n
c
tio
n
()
fr
to
g
et
th
9
e
ac
cu
r
ate
s
k
eleto
n
[
2
6
]
:
(
)
1
(
)
(
)
f
r
G
r
d
t
r
(
1
)
w
h
er
e
()
Gr
r
ep
r
esen
ts
a
Ga
u
s
s
ian
k
er
n
el
f
u
n
ct
io
n
,
()
d
t
r
r
ep
r
esen
ts
d
is
ta
n
ce
tr
a
n
s
f
o
r
m
w
h
ic
h
is
t
h
e
d
is
tan
ce
f
r
o
m
an
i
n
ter
io
r
p
o
in
t
r
to
th
e
n
ea
r
est
b
o
u
n
d
ar
y
p
o
i
n
t,
is
its
s
tan
d
ar
d
co
v
ar
ian
ce
,
an
d
is
th
e
co
n
v
o
lu
tio
n
o
p
er
ato
r
.
We
co
m
p
u
te
t
h
e
g
r
ad
ien
t
of
()
fr
as
f
o
llo
w
:
00
(
,
)
(
,
)
ff
u
v
f
xy
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8776
IJ
-
I
C
T
Vo
l.
8
,
No
.
1
,
A
p
r
il
2
0
1
9
:
1
–
12
4
T
h
e
is
o
tr
o
p
ic
d
if
f
u
s
io
n
of
()
fr
is
p
er
f
o
r
m
ed
:
2
2
2
2
2
2
(
)
(
)
(
)
(
)
x
x
y
y
x
y
du
u
u
u
f
f
f
dt
dv
u
u
v
f
f
f
dt
(
3)
w
h
er
e
,
uv
ar
e
t
w
o
co
m
p
o
n
e
n
ts
of
th
e
d
if
f
u
s
ed
g
r
ad
ien
t
v
ec
to
r
f
ield
.
x
f
an
d
y
f
ar
e
th
e
t
w
o
co
m
p
o
n
en
t
s
of
()
fr
.
I
n
itializi
n
g
,
uv
w
i
th
00
,
uv
in
(
2
)
,
th
e
p
ar
tial
d
if
f
er
en
tia
l
(
3
)
can
be
s
o
lv
ed
iter
ati
v
el
y
by
f
in
ite
d
if
f
er
e
n
ce
tech
n
iq
u
e.
T
h
e
is
o
tr
o
p
ic
d
if
f
u
s
io
n
m
a
k
es
th
e
v
ec
to
r
s
of
()
fr
p
r
o
p
ag
ate
to
w
ar
d
s
to
th
e
o
b
j
ec
t
ce
n
tr
e,
an
d
th
e
in
ter
s
ec
tio
n
s
of
v
ec
to
r
s
d
eter
m
i
n
e
th
e
ac
t
u
al
lo
ca
tio
n
of
th
e
s
k
eleto
n
p
o
in
ts
.
T
h
en
,
we
co
n
s
id
er
th
e
i
n
itial
g
r
ad
ien
t
v
ec
to
r
f
ield
()
g
v
f
r
:
(
)
(
(
)
(
))
rr
g
vf
r
I
r
I
r
rr
(
4
)
w
h
er
e
()
Ir
is
t
h
e
in
te
n
s
it
y
v
al
u
e
at
r
,
an
d
r
r
ep
r
esen
ts
o
n
e
of
th
e
ei
g
h
t
i
m
m
ed
iate
n
eig
h
b
o
r
s
of
r
.
T
h
e
SSM
v
al
u
e
at
ea
c
h
p
o
in
t
i
n
d
icate
s
t
h
e
p
r
o
b
ab
ilit
y
of
b
ei
n
g
a
s
k
eleto
n
p
o
i
n
t.
T
h
e
h
i
g
h
e
r
v
alu
e
at
a
p
o
in
t,
th
e
m
o
r
e
p
r
o
b
ab
le
th
is
p
o
in
t
is
a
s
k
ele
to
n
p
o
in
t.
We
co
m
p
u
te
t
h
e
SSM
v
al
u
e
b
y
:
(
(
)
)
(
)
(
)
(
)
m
ax
(
0
,
)
r
N
r
g
vf
r
r
r
S
S
M
r
rr
(
5
)
w
h
er
e
()
Nr
is
t
h
e
s
et
of
t
h
e
ei
g
h
t
i
m
m
ed
iate
n
ei
g
h
b
o
u
r
s
of
r
.
2
.
2
.
2.
Ref
ined
SSM
v
a
lue
W
h
en
we
h
a
v
e
b
ee
n
co
m
p
u
te
d
th
e
v
al
u
e
of
SS
M,
u
s
e
n
o
n
-
m
ax
i
m
a
l
s
u
p
p
r
ess
io
n
[
2
7
]
alg
o
r
ith
m
to
o
b
tain
SSM
r
ef
i
n
ed
ed
g
e.
T
h
e
n
o
n
-
m
ax
i
m
al
s
u
p
p
r
ess
io
n
alg
o
r
ith
m
is
d
escr
ib
ed
as
f
o
llo
w
s
:
Fig
u
r
e
3
(
a)
s
h
o
w
s
th
e
e
ig
h
t
d
i
r
ec
tio
n
s
of
p
o
i
n
t
(
,
)
P
x
y
w
it
h
i
n
t
h
e
3
×3
r
eg
io
n
.
In
F
ig
u
r
e
3
(
b
)
,
th
e
q
u
ad
r
an
g
le
is
f
o
r
m
ed
by
co
n
n
ec
ti
n
g
th
e
e
ig
h
t
d
ir
ec
tio
n
s
of
p
o
in
t
(
,
)
P
x
y
.
T
h
en
,
we
u
s
e
t
h
e
d
ir
ec
tio
n
of
()
g
vf
P
to
m
a
k
e
a
s
tr
ai
g
h
t
lin
e,
w
h
ic
h
in
ter
s
ec
t
w
it
h
t
h
e
q
u
ad
r
an
g
le
at
p
o
in
t
(
,
)
xy
an
d
(
,
)
xy
.
If
(
,
)
(
,
)
&
(
,
)
(
,
)
S
S
M
x
y
S
S
M
x
y
S
S
M
x
y
S
S
M
x
y
,
th
en
t
h
e
p
o
in
t
(
,
)
P
x
y
w
il
l
be
r
etain
ed
.
If
n
o
t,
th
e
p
o
i
n
t
(
,
)
P
x
y
w
il
l
be
d
elete
d
.
W
h
en
all
p
o
in
ts
h
av
e
b
ee
n
u
tili
ze
d
t
h
e
n
o
n
-
m
ax
i
m
a
l
s
u
p
p
r
ess
io
n
alg
o
r
it
h
m
,
we
w
o
u
ld
g
et
t
h
e
r
ef
i
n
ed
SSM
v
al
u
e.
3
2
1
4
P
8
7
6
5
(
a)
P
=
(
x
,
y
)
(
x
'
,
y
'
)
(
x
"
,
y
"
)
(
b
)
Fig
u
r
e
3.
(
a)
T
he
eig
h
t
d
ir
ec
ti
o
n
s
of
p
o
in
t
,
(
b
)
T
h
e
q
u
ad
r
an
g
le
f
o
r
m
ed
by
co
n
n
ec
ti
n
g
t
h
e
ei
g
h
t
d
ir
ec
tio
n
s
of
p
o
in
t
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
A
n
o
ve
l
s
ke
tch
-
b
a
s
ed
3D
mo
d
el
r
etri
ev
a
l
a
p
p
r
o
a
ch
b
a
s
ed
on
s
ke
leto
n
(
Jin
g
Zh
a
n
g
)
5
2
.
2
.
3.
Select
io
n
of
cr
it
ica
l
po
i
nts
Fo
r
th
e
r
ef
i
n
ed
SSM
v
al
u
e,
we
se
lect
t
h
e
cr
itical
p
o
in
t
s
w
it
h
t
h
e
lo
w
es
t
v
al
u
e
of
th
e
g
r
ad
ien
t
m
ag
n
it
u
d
e
(
)
(
)
G
r
d
t
r
.
Ou
r
cr
itical
p
o
in
ts
co
r
r
esp
o
n
d
to
s
ig
n
i
f
ican
t
v
is
u
al
p
ar
ts
of
th
e
o
b
j
ec
t.
T
h
er
ef
o
r
e,
th
e
o
b
tain
ed
s
k
eleto
n
co
n
tai
n
s
b
r
an
c
h
es
r
e
p
r
esen
tin
g
all
s
ig
n
i
f
ica
n
t
v
i
s
u
al
p
ar
ts
.
No
tice
th
at
in
th
e
d
e
f
i
n
itio
n
of
th
e
cr
itical
p
o
in
t,
we
a
ls
o
in
c
lu
d
e
t
h
e
en
d
p
o
in
ts
,
b
ec
au
s
e
th
o
s
e
p
o
in
ts
u
s
u
all
y
do
n
o
t
h
av
e
m
i
n
i
m
u
m
g
r
ad
ien
t
m
a
g
n
i
tu
d
e.
If
th
e
y
ar
e
n
o
t
s
elec
ted
,
th
e
s
k
eleto
n
b
r
an
ch
e
s
m
a
y
be
s
h
o
r
te
n
ed
.
2
.
2
.
4.
Sk
elet
o
n
t
ra
ce
W
e
co
n
s
id
er
th
e
cr
itical
p
o
in
ts
’
E
u
clid
ea
n
d
i
s
ta
n
ce
m
a
tr
ix
to
r
ep
r
esen
t
a
n
u
n
d
ir
ec
ted
w
ei
g
h
ted
g
r
ap
h
.
E
ac
h
cr
itical
p
o
in
t
ch
o
o
s
es
t
h
e
to
p
3
m
i
n
i
m
u
m
E
u
cl
i
d
ea
n
d
is
ta
n
ce
w
it
h
o
t
h
er
p
o
in
t
s
in
t
h
e
p
r
o
c
ess
o
f
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
[
2
0
]
.
T
h
e
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
ad
d
s
ed
g
e
s
b
y
w
ei
g
h
t
asce
n
d
in
g
o
r
d
er
,
w
h
ic
h
f
o
r
m
s
a
tr
e
e
th
at
i
n
clu
d
es
e
v
er
y
v
er
tex
.
T
h
e
to
tal
w
ei
g
h
t
o
f
all
th
e
ed
g
e
s
in
th
e
tr
ee
is
m
in
i
m
ized
.
B
elo
w
ar
e
th
e
s
tep
s
o
f
co
n
n
ec
ti
n
g
cr
itical
p
o
in
ts
u
s
i
n
g
Kr
u
s
k
al
’
s
alg
o
r
it
h
m
.
Ass
u
m
e
(
,
)
G
V
E
r
ep
r
esen
t
g
r
ap
h
w
h
ic
h
f
o
r
m
ed
by
cr
itical
p
o
in
t
s
’
E
u
clid
ea
n
d
i
s
tan
ce
.
W
h
er
e
V
is
t
h
e
cr
itica
l
p
o
in
t
s
s
e
t
of
th
e
g
r
ap
h
G
,
an
d
E
is
th
e
ed
g
e
s
et
of
t
h
e
cr
itical
p
o
in
ts
’
E
u
clid
ea
n
d
is
tan
ce
.
W
e
s
et
th
e
n
u
m
b
er
of
cr
itica
l
p
o
in
ts
is
N
,
an
d
s
o
r
t
all
th
e
ed
g
e
s
by
w
eig
h
t
a
s
ce
n
d
i
n
g
o
r
d
er
.
1.
Step
1
.
We
d
ef
in
e
a
s
et
of
N
i
n
d
ep
en
d
en
t
v
er
tices.
Si
n
ce
N
cr
itical
p
o
in
ts
n
ee
d
be
co
n
n
e
ct
ed
,
we
co
n
s
id
er
th
e
N
cr
itical
p
o
in
ts
s
ep
ar
atel
y
.
2.
Step
2
.
We
ch
o
o
s
e
t
he
ed
g
e
by
w
ei
g
h
t
a
s
ce
n
d
i
n
g
o
r
d
er
.
If
t
h
e
ed
g
e
of
t
w
o
v
er
tices
s
at
is
f
ies
in
d
i
f
f
er
en
t
v
er
tex
s
e
ts
,
we
ad
d
th
is
ed
g
e
to
th
e
m
in
i
m
u
m
s
p
an
n
i
n
g
tr
ee
’s
ed
g
e
s
et,
a
n
d
m
er
g
e
t
h
e
t
w
o
d
if
f
er
en
t
v
er
tex
s
e
ts
i
n
to
o
n
e
v
er
te
x
s
et.
If
n
o
t,
we
w
o
u
ld
co
n
s
id
er
t
h
e
v
er
tices
of
n
e
x
t
ed
g
e.
3.
Step
3
.
R
ep
ea
t
s
tep
2
u
n
til
all
t
h
e
v
er
tic
es
ar
e
in
t
h
e
s
a
m
e
v
er
t
ex
s
et.
As
s
h
o
w
n
in
Fi
g
u
r
e
4
(
a)
,
it
is
an
ex
a
m
p
le
to
e
x
p
lai
n
o
u
r
ap
p
r
o
ac
h
w
h
ic
h
i
n
cl
u
d
es
7
cr
itica
l
p
o
in
ts
to
u
s
e
th
e
Kr
u
s
k
a
l’
s
al
g
o
r
ith
m
.
In
F
ig
u
r
e
4
(
b
)
,
th
e
E
u
cl
id
ea
n
d
is
tan
ce
b
et
w
ee
n
7
cr
itical
p
o
in
ts
r
ep
r
ese
n
ts
an
u
n
d
ir
ec
ted
co
m
p
le
te
g
r
ap
h
.
In
Fi
g
u
r
e
4
(
c)
,
it
’
s
th
e
m
in
i
m
u
m
s
p
an
n
in
g
tr
ee
of
7
cr
itical
p
o
in
ts
w
h
ic
h
u
s
e
th
e
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
to
cr
ea
te.
A
B
C
D
E
F
G
(
a)
5
9
1
1
8
8
5
6
7
7
1
2
A
B
C
D
E
F
G
1
2
9
1
0
(
b
)
5
9
5
6
7
7
A
B
C
D
E
F
G
(
c)
Fig
u
r
e
4
.
(
a)
An
ex
a
m
p
le
i
n
cl
u
d
in
g
7
cr
itical
p
o
in
ts
,
(
b
)
T
he
E
u
clid
ea
n
d
is
ta
n
ce
b
et
w
ee
n
7
cr
itical
p
o
in
ts
,
(
c)
T
h
e
m
in
i
m
u
m
s
p
an
n
i
n
g
tr
ee
of
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
2
.
3
.
H
is
t
o
g
ra
m
F
ea
t
ure
Co
m
pa
r
is
on
Ou
r
h
is
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
alg
o
r
ith
m
co
n
s
is
t
s
of
h
i
s
to
g
r
a
m
f
ea
t
u
r
e
ex
tr
ac
tio
n
an
d
s
k
eleto
n
d
is
tan
ce
co
m
p
u
tatio
n
al
g
o
r
ith
m
.
T
h
e
h
is
to
g
r
a
m
f
ea
t
u
r
e
ex
t
r
ac
tio
n
alg
o
r
ith
m
ad
o
p
ts
th
e
r
ad
ii
of
th
e
d
is
k
s
at
s
k
eleto
n
p
o
in
ts
a
n
d
th
e
len
g
t
h
s
of
s
k
ele
to
n
b
r
an
ch
es
to
ex
tr
ac
t
th
e
h
i
s
to
g
r
a
m
f
ea
t
u
r
e.
T
h
e
r
ad
ii
of
th
e
d
is
k
s
an
d
len
g
t
h
s
of
s
k
eleto
n
b
r
an
c
h
es
ar
e
in
v
ar
ia
n
t
u
n
d
er
th
e
e
n
v
ir
o
n
m
en
t
of
n
o
n
-
r
i
g
id
tr
an
s
f
o
r
m
at
io
n
[
2
8
]
w
h
ic
h
w
o
u
ld
h
elp
to
g
et
h
ig
h
-
p
r
ec
is
io
n
r
etr
iev
a
l
r
esu
lt.
A
d
d
itio
n
a
ll
y
,
th
e
s
k
eleto
n
d
is
tan
ce
co
m
p
u
tatio
n
al
g
o
r
ith
m
co
m
p
ar
es
t
h
e
s
i
m
ilar
it
y
b
et
w
e
en
t
w
o
s
k
eleto
n
s
u
s
i
n
g
th
e
h
is
to
g
r
a
m
f
ea
tu
r
e
m
atr
ix
of
s
k
ele
to
n
en
d
p
o
in
ts
,
w
it
h
r
elativ
el
y
les
s
q
u
an
tit
y
t
h
a
n
s
k
eleto
n
p
o
in
ts
w
h
ic
h
lead
s
to
a
lo
w
er
co
m
p
u
tatio
n
al
co
m
p
le
x
i
t
y
.
2
.
3
.
1.
H
is
t
o
g
ra
m
F
ea
t
ure
E
x
t
ra
ct
io
n
We
d
ef
in
e
a
2D
d
ataset
{
(
,
)
1
,
2
,
.
.
.
,
}
ii
S
a
b
i
T
,
an
d
th
e
(
,
)
H
S
X
r
ep
r
esen
ts
th
e
2D
h
is
to
g
r
a
m
m
atr
i
x
of
S
,
w
h
er
e
=
{
(
,
)
0
,
1
,
.
.
.
,
;
0
,
1
,
.
.
.
,
}
ij
X
x
y
i
m
j
n
is
t
h
e
h
is
to
g
r
a
m
p
ar
a
m
et
er
,
w
it
h
co
n
s
tr
ain
ts
-
1
-
1
,
i
i
i
i
x
x
y
y
,
1
[
,
]
im
a
x
x
,
1
[
,
]
jn
b
y
y
.
Def
in
e
t
h
e
2D
h
i
s
to
g
r
a
m
m
atr
ix
(
,
)
H
S
X
as
f
o
llo
w
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8776
IJ
-
I
C
T
Vo
l.
8
,
No
.
1
,
A
p
r
il
2
0
1
9
:
1
–
12
6
11
12
1
21
22
2
12
(
,
)
=
n
n
m
m
mn
h
h
h
h
h
h
H
S
X
h
h
h
(
6
)
w
h
er
e
ij
h
r
ep
r
esen
ts
t
h
e
n
u
m
b
e
r
of
S
d
ataset
ele
m
e
n
ts
d
is
tr
ib
u
ti
n
g
in
t
h
e
r
ec
tan
g
u
lar
ar
e
a
(
1
,
ii
xx
,
1
,
jj
yy
)
,
,
ij
xy
u
s
u
all
y
is
ar
ith
m
etic
s
eq
u
e
n
ce
or
g
eo
m
etr
ic
s
eq
u
e
n
ce
.
As
s
h
o
w
n
in
Fi
g
u
r
e
5,
th
e
p
o
i
n
t
Q
r
ep
r
esen
ts
th
e
s
k
eleto
n
ce
n
tr
e
w
h
ic
h
is
th
e
s
k
eleto
n
p
o
in
t
w
it
h
r
ad
iu
s
of
t
h
e
m
a
x
i
m
a
l
d
is
k
,
a
n
d
(
1
,
2
,
.
.
,
)
i
v
i
T
is
th
e
s
et
of
s
k
eleto
n
p
o
in
t
s
.
T
h
en
we
u
s
e
f
ea
tu
r
e
d
ata
s
et
Q
P
to
r
ep
lace
th
e
ab
o
v
e
2D
d
atase
t
S
:
{
(
)
(
)
(
(
,
)
,
(
)
)
}
(
1
,
2
,
.
.
.
,
)
Q
Q
i
Q
i
i
i
P
p
v
p
v
s
k
e
Q
v
R
v
i
T
(
7
)
w
h
er
e
(
,
)
i
s
k
e
Q
v
is
th
e
s
h
o
r
tes
t
p
ath
f
r
o
m
s
k
eleto
n
ce
n
tr
e
Q
to
s
k
eleto
n
p
o
in
t
i
v
.
()
i
Rv
r
ep
r
esen
ts
t
h
e
r
ad
i
i
of
d
is
k
s
of
s
k
ele
to
n
p
o
in
t
i
v
.
i
v
is
th
e
co
r
r
esp
o
n
d
in
g
p
o
in
t
of
i
v
at
th
e
co
n
to
u
r
.
()
Qi
pv
r
ep
r
esen
ts
th
e
r
elat
iv
e
d
is
tan
ce
f
r
o
m
s
k
eleto
n
ce
n
tr
e
Q
to
co
n
to
u
r
p
o
in
t
i
v
.
Fig
u
r
e
5.
T
h
e
r
ad
iu
s
of
th
e
m
a
x
i
m
al
d
is
k
an
d
co
n
to
u
r
p
o
in
t
Fo
r
th
e
2D
h
is
to
g
r
a
m
f
ea
t
u
r
e
m
atr
i
x
(
,
)
Q
H
P
X
,
X
is
th
e
co
r
r
esp
o
n
d
in
g
h
i
s
to
g
r
a
m
p
ar
a
m
eter
.
{
(
,
)
,
}
(
0
,
1
,
.
.
.
,
;
0
,
1
,
.
.
.
,
)
i
i
j
i
j
X
x
y
x
l
y
r
j
i
m
j
n
(
8
)
T
h
e
d
eter
m
i
n
atio
n
s
of
p
ar
a
m
e
ter
s
,
,
,
l
r
m
n
ar
e
u
n
d
er
th
e
co
n
s
tr
ai
nt
of
r
ad
ii
of
d
is
k
s
an
d
t
h
e
len
g
t
h
s
of
s
k
e
leto
n
b
r
an
c
h
es.
As
s
h
o
w
n
in
Fi
g
u
r
e
6,
we
co
n
s
id
er
th
e
en
d
p
o
in
t
A
as
th
e
b
a
s
e
r
esear
ch
p
o
in
t,
t
h
e
d
ef
i
n
itio
n
of
A
P
is
s
i
m
ilar
to
Q
P
.
T
h
e
(
,
)
A
H
P
X
r
ep
r
esen
ts
t
h
e
h
is
to
g
r
a
m
f
ea
t
u
r
e
m
atr
i
x
.
Fig
u
r
e
6.
T
h
e
s
h
o
r
test
p
ath
of
s
k
eleto
n
en
d
p
o
in
t
s
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
A
n
o
ve
l
s
ke
tch
-
b
a
s
ed
3D
mo
d
el
r
etri
ev
a
l
a
p
p
r
o
a
ch
b
a
s
ed
on
s
ke
leto
n
(
Jin
g
Zh
a
n
g
)
7
T
h
e
Fig
u
r
e
7
s
h
o
w
s
th
e
2D
h
i
s
to
g
r
a
m
of
s
k
eleto
n
en
d
p
o
in
t
A
g
en
er
ated
by
t
h
e
le
n
g
th
of
a
ll
cu
r
v
e
s
d
is
tr
ib
u
ti
n
g
in
ea
c
h
r
ec
ta
n
g
u
l
ar
ar
ea
.
T
h
e
(
,
)
i
s
k
e
A
v
is
th
e
ab
s
cis
s
a
wh
ich
r
ep
r
esen
t
s
t
h
e
s
h
o
r
tes
t
p
a
th
f
r
o
m
th
e
e
n
d
p
o
in
t
A
to
s
k
e
leto
n
p
o
i
n
t
i
v
.
T
h
e
()
i
Rv
is
th
e
o
r
d
in
ate
w
h
ic
h
r
ep
r
esen
ts
t
h
e
r
ad
iu
s
of
d
i
s
k
of
s
k
e
leto
n
p
o
in
t
i
v
.
Fig
u
r
e
7
.
T
h
e
2D
h
is
to
g
r
a
m
of
s
k
eleto
n
en
d
p
o
in
t
A
2
.
3
.
2.
Sk
elet
o
n
Di
s
t
a
nce
Co
m
p
uta
t
io
n
We
d
ef
in
e
t
w
o
s
k
eleto
n
s
12
=
(
v
,
v
,
.
.
.
,
v
)
m
G
an
d
12
=
(
v
,
v
,
.
.
.
,
v
)
n
G
,
w
h
er
e
v
i
an
d
v
j
r
ep
r
esen
t
en
d
p
o
in
ts
f
r
o
m
d
i
f
f
er
en
t
s
k
eleto
n
s
.
W
it
h
t
h
e
h
is
to
g
r
a
m
f
ea
t
u
r
e
m
a
tr
ix
()
i
Hk
an
d
()
j
Hk
of
s
k
el
eto
n
en
d
p
o
in
ts
v
i
an
d
v
j
,
we
g
et
t
h
e
d
is
ta
n
ce
b
et
w
ee
n
2D
h
i
s
to
g
r
a
m
of
t
w
o
s
k
eleto
n
en
d
p
o
in
ts
:
1
(
v
,
v
)
(
)
(
)
mn
i
j
k
i
j
k
d
w
H
k
H
k
(
9
)
w
h
er
e
1
(
)
+
(
)
co
s
(
)
2
=
(
)
+
(
)
co
s
(
)
2
ij
k
mn
ij
k
H
k
H
k
Z
w
H
k
H
k
Z
(
1
0
)
2
m
ax
(
(
)
+
(
))
ij
k
Z
H
k
H
k
(
1
1
)
s
atis
f
y
in
g
(
)
+
(
)
1
ij
H
k
H
k
Z
.
W
h
en
we
co
m
p
ar
e
t
h
e
s
i
m
ilar
it
y
b
et
w
ee
n
t
wo
s
k
e
leto
n
s
,
ch
o
o
s
e
o
n
e
of
th
e
s
k
e
l
eto
n
s
as
a
b
en
ch
m
ar
k
,
a
n
d
co
m
p
u
te
all
t
h
e
d
is
ta
n
ce
w
it
h
th
e
o
t
h
er
s
k
el
eto
n
en
d
p
o
in
ts
.
1
1
1
2
1
2
1
2
2
2
12
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
(
,
)
n
n
m
m
m
n
d
v
v
d
v
v
d
v
v
d
v
v
d
v
v
d
v
v
D
G
G
d
v
v
d
v
v
d
v
v
(
1
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8776
IJ
-
I
C
T
Vo
l.
8
,
No
.
1
,
A
p
r
il
2
0
1
9
:
1
–
12
8
In
th
e
all
e
n
d
p
o
in
ts
d
is
tan
c
e
m
atr
i
x
(
,
)
D
G
G
,
we
s
elec
t
t
h
e
m
in
i
m
u
m
v
al
u
e
of
each
lin
e
12
m
i
n
(
,
)
m
i
n
(
,
)
m
i
n
(
,
)
(
1
,
2
,
.
.
.
,
)
j
j
m
j
d
v
v
d
v
v
d
v
v
j
n
，
，
，
to
co
m
p
u
te
t
h
e
d
is
ta
n
ce
b
et
w
ee
n
t
w
o
s
k
eleto
n
s
.
T
h
en
t
h
e
d
is
ta
n
ce
of
t
w
o
s
k
eleto
n
s
is
e
x
p
r
ess
ed
as
:
1
d
i
s
(
,
)
m
i
n
(
,
)
,
1
,
2
,
.
.
.
,
m
ij
j
i
G
G
d
v
v
j
n
(
1
3
)
3.
RE
SU
L
T
S
AND
AN
AL
Y
SI
S
We
i
m
p
le
m
e
n
t
o
u
r
s
k
etc
h
-
b
as
ed
3D
m
o
d
el
r
etr
iev
al
m
et
h
o
d
in
C
++
u
n
d
er
W
in
d
o
w
s
.
As
s
h
o
w
n
in
Fig
u
r
e
8,
T
h
e
lef
t
s
id
e
of
t
h
e
i
n
ter
f
ac
e
is
a
ca
n
v
a
s
f
o
r
s
k
e
tch
in
g
t
h
e
m
o
d
el.
T
h
e
u
s
er
ca
n
e
r
ase
an
d
m
o
d
if
y
if
th
e
y
do
n
o
t
s
ati
s
f
y
w
i
th
t
h
e
d
r
a
w
n
s
k
etc
h
.
T
h
e
r
ig
h
t
s
id
e
is
d
is
p
la
y
i
n
g
p
ag
e
f
o
r
th
e
r
etr
iev
ed
3D
m
o
d
els
w
it
h
a
r
elev
an
t
J
P
E
G
i
m
ag
e.
T
h
e
u
s
er
can
click
t
h
e
b
lu
e
b
u
tto
n
to
d
o
w
n
lo
ad
th
e
co
r
r
esp
o
n
d
in
g
3D
m
o
d
el.
T
h
e
s
y
s
te
m
co
n
s
is
ts
of
o
f
f
-
li
n
e
f
ea
t
u
r
e
ex
tr
ac
tio
n
a
n
d
on
-
lin
e
r
etr
iev
al
p
r
o
ce
s
s
es.
In
th
e
o
f
f
-
li
n
e
p
r
o
ce
s
s
,
th
e
f
ea
t
u
r
es
ar
e
e
x
tr
a
cted
in
a
PC
w
it
h
a
P
en
ti
u
m
III
8
0
0
MH
z
C
P
U
an
d
GeFo
r
ce
2
MX
v
id
eo
ca
r
d
.
In
th
e
on
-
l
in
e
p
r
o
ce
s
s
,
t
h
e
r
etr
ie
v
al
s
y
s
te
m
co
n
s
is
ts
by
a
PC
w
it
h
an
I
n
tel
Xeo
n
C
P
U
E
5
5
2
0
@
2
.
2
7
GHz
an
d
1
2
.
0
GB
of
R
A
M.
Ou
r
s
k
etc
h
-
b
a
s
ed
3D
m
o
d
el
r
etr
iev
al
b
en
ch
m
ar
k
is
b
u
ilt
on
th
e
well
-
k
n
o
w
n
Natio
n
al
T
aiw
a
n
U
n
iv
er
s
it
y
(
NT
U)
[
2
9
]
d
atab
ase
an
d
th
e
latest
co
llect
io
n
of
h
u
m
an
s
k
etc
h
e
s
.
Fig
u
r
e
8.
Ou
r
s
k
etc
h
-
b
ased
3D
m
o
d
el
r
etr
iev
al
s
y
s
te
m
3.
1
.
Sk
elet
o
n
E
x
t
ra
ct
io
n
by
ASSM
T
h
e
p
r
o
p
o
s
ed
A
SS
M
m
et
h
o
d
is
ab
le
to
co
m
p
u
te
s
k
ele
to
n
b
r
a
n
ch
es
in
all
s
i
g
n
i
f
ica
n
t
v
is
u
al
p
ar
ts
.
Fig
u
r
e
9
s
h
o
w
s
lea
f
s
k
eleto
n
e
x
tr
ac
tio
n
p
r
o
ce
s
s
by
ASSM
ap
p
r
o
ac
h
.
As
s
h
o
w
n
in
Fi
g
u
r
e
9
(
a)
,
it’
s
t
h
e
o
r
ig
i
n
al
leaf
i
m
a
g
e.
In
Fi
g
u
r
e
9
(
b
)
,
s
h
o
w
s
t
h
e
co
m
p
u
tatio
n
of
th
e
S
SM
v
alu
e
a
f
ter
is
o
tr
o
p
ic
d
if
f
u
s
io
n
on
th
e
g
r
ad
ie
n
t
v
ec
to
r
f
ield
.
In
Fig
u
r
e
9
(
c)
,
t
h
e
SSM
v
a
lu
e
is
r
ef
in
ed
by
t
h
e
n
o
n
-
m
a
x
i
m
al
s
u
p
p
r
ess
io
n
alg
o
r
ith
m
.
In
Fi
g
u
r
e
9
(
d
)
,
th
e
cr
itical
p
o
in
t
s
s
et
e
x
tr
ac
ted
f
r
o
m
th
e
SS
M
r
ef
in
ed
v
al
u
e.
In
Fi
g
u
r
e
9
(
e)
,
s
h
o
w
s
th
e
f
in
al
s
k
eleto
n
co
n
n
ec
ted
by
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
.
(
a)
(
b
)
(
c)
(
d
)
(
e)
Fig
u
r
e
9
.
I
llu
s
tr
atio
n
of
leaf
s
k
eleto
n
ex
tr
ac
tio
n
p
r
o
ce
s
s
by
ASSM
ap
p
r
o
ac
h
(
a)
th
e
o
r
ig
in
al
lea
f
i
m
ag
e
,
(
b
)
s
h
o
w
s
t
h
e
SS
M
v
al
u
e,
(
c)
th
e
SSM
v
al
u
e
is
r
ef
i
n
ed
by
th
e
non
-
m
a
x
i
m
al
s
u
p
p
r
ess
io
n
alg
o
r
it
h
m
,
(
d
)
th
e
cr
itical
p
o
in
t
s
et
ex
tr
ac
ted
f
r
o
m
(
c)
,
(
e)
th
e
f
in
a
l
s
k
e
leto
n
co
n
n
ec
ted
by
Kr
u
s
k
al
'
s
al
g
o
r
ith
m
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
I
C
T
I
SS
N:
2252
-
8776
A
n
o
ve
l
s
ke
tch
-
b
a
s
ed
3D
mo
d
el
r
etri
ev
a
l
a
p
p
r
o
a
ch
b
a
s
ed
on
s
ke
leto
n
(
Jin
g
Zh
a
n
g
)
9
3.
2
.
Sk
elet
o
n
M
a
t
ching
B
a
s
e
d
on
H
is
t
o
g
ra
m
F
ea
t
ure
In
o
r
d
er
to
s
h
o
w
t
h
e
ef
f
ec
ti
v
en
es
s
of
t
h
e
p
r
o
p
o
s
ed
h
is
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
alg
o
r
ith
m
,
we
test
ed
th
r
ee
s
k
e
leto
n
m
atc
h
i
ng
e
x
p
er
i
m
e
n
ts
b
et
w
ee
n
s
k
e
tch
an
d
2D
v
ie
w
of
3D
m
o
d
el.
F
ig
u
r
e
1
0
s
h
o
w
s
th
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
a
cat
an
d
a
n
o
th
er
d
e
f
o
r
m
ed
cat
.
I
t
d
e
m
o
n
s
tr
ates
t
h
at
o
u
r
h
i
s
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
alg
o
r
it
h
m
ca
n
r
ea
lize
th
e
s
k
e
leto
n
m
atch
in
g
of
d
ef
o
r
m
ed
o
b
j
ec
t
.
Fig
u
r
e
1
1
s
h
o
w
s
th
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
t
h
e
p
er
s
o
n
s
w
i
th
d
if
f
er
e
n
t
n
u
m
b
er
s
of
ar
m
s
.
It
ill
u
s
tr
ate
s
t
h
at
o
u
r
h
i
s
to
g
r
a
m
f
ea
t
u
r
e
co
m
p
ar
is
o
n
al
g
o
r
ith
m
w
o
r
k
s
co
r
r
ec
tl
y
if
o
b
j
ec
t
p
ar
ts
a
r
e
s
ig
n
i
f
ica
n
tl
y
al
ter
ed
.
Fig
u
r
e
1
2
s
h
o
w
s
th
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
an
e
lep
h
an
t
a
nd
an
o
t
h
er
elep
h
a
n
t
w
it
h
a
s
tick
.
It
d
e
m
o
n
s
tr
ates
th
at
o
u
r
m
atc
h
in
g
p
r
o
ce
s
s
h
as
s
tr
o
n
g
p
er
f
o
r
m
a
n
c
e
w
h
en
t
h
e
co
m
p
ar
ed
o
b
j
ec
ts
h
av
e
t
h
e
r
ed
u
n
d
an
t
p
ar
ts
.
Fig
u
r
e
1
0.
T
h
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
a
cat
an
d
an
o
t
h
er
d
ef
o
r
m
ed
cat
Fig
u
r
e
1
1.
T
h
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
th
e
p
er
s
o
n
s
w
it
h
d
if
f
er
e
n
t
n
u
m
b
er
s
of
ar
m
s
Fig
u
r
e
12.
T
h
e
co
r
r
esp
o
n
d
en
ce
b
et
w
ee
n
an
elep
h
a
n
t
a
n
d
an
o
t
h
er
elep
h
an
t
w
i
th
a
s
tick
3.
3
.
Co
m
pa
ri
s
o
n
w
it
h
o
t
her
a
pp
ro
a
ches
We
co
n
s
id
er
ed
class
ical
P
r
ec
i
s
io
n
an
d
R
ec
all
m
etr
ics
av
er
a
g
ed
o
v
er
th
e
s
et
of
p
r
o
ce
s
s
ed
q
u
er
ies
[
3
0
]
to
m
ea
s
u
r
e
th
e
r
etr
ie
v
al
e
f
f
ec
t
i
v
en
e
s
s
.
R
ec
all
m
ea
s
u
r
es
t
h
e
a
b
ilit
y
of
th
e
s
y
s
te
m
to
r
etr
iev
e
all
m
o
d
els
t
h
at
ar
e
r
elev
an
t.
P
r
ec
is
io
n
m
ea
s
u
r
es
t
h
at
t
h
e
ab
ilit
y
of
t
h
e
s
y
s
te
m
to
r
etr
i
ev
e
o
n
l
y
m
o
d
els
t
h
at
ar
e
r
elev
an
t.
T
h
e
y
ar
e
d
ef
in
ed
as:
(
1
4
)
(
1
5
)
T
o
h
av
e
a
co
m
p
r
eh
e
n
s
iv
e
e
v
alu
atio
n
o
f
o
u
r
al
g
o
r
ith
m
,
we
f
u
r
t
h
er
p
r
o
v
id
e
th
e
r
es
u
lt
s
f
o
r
o
th
er
p
er
f
o
r
m
a
n
ce
m
etr
ic
s
i
n
clu
d
i
n
g
Nea
r
est
Nei
g
h
b
o
u
r
(
NN)
,
F
ir
s
t
T
ier
(
FT)
,
Seco
n
d
T
ier
(
S
T
)
,
E
-
m
ea
s
u
r
e
(
E
)
,
Dis
co
u
n
ted
C
u
m
u
lati
v
e
Gain
(
DC
G)
an
d
Av
er
ag
e
P
r
ec
is
io
n
(
A
P
)
.
T
h
e
m
ea
n
i
n
g
o
f
th
e
ab
o
v
e
p
er
f
o
r
m
a
n
ce
m
etr
ics
is
a
s
f
o
llo
w
s
[
3
0
]
.
N
N
m
ea
s
u
r
e
s
t
h
e
p
er
ce
n
ta
g
e
o
f
th
e
c
lo
s
est
m
atc
h
e
s
t
h
at
ar
e
r
elev
an
t
m
o
d
el
s
.
FT
r
ep
r
esen
ts
h
o
w
m
u
ch
p
er
ce
n
ta
g
e
o
f
a
clas
s
h
a
s
b
ee
n
r
etr
iev
e
d
am
o
n
g
t
h
e
to
p
C
lis
t,
w
h
er
e
C
is
t
h
e
ca
r
d
in
alit
y
o
f
th
e
r
elev
a
n
t c
la
s
s
o
f
t
h
e
q
u
e
r
y
s
k
e
tch
.
I
t d
ef
i
n
es a
s
:
(
1
)
r
e
l
e
v
a
n
t
c
o
r
r
e
c
t
l
y
r
e
t
r
i
e
v
e
d
FT
t
o
p
C
r
e
t
r
i
e
v
e
d
(1
6
)
ST
r
e
p
r
esen
ts
h
o
w
m
u
c
h
p
er
ce
n
tag
e
of
a
class
h
as
b
ee
n
r
etr
iev
ed
a
m
o
n
g
th
e
to
p
2
(
1
)
C
lis
t
,
w
h
er
e
C
h
as
t
h
e
s
a
m
e
m
ea
n
i
n
g
w
it
h
FT
m
e
tr
ic
.
2
(
1
)
r
e
l
e
v
a
n
t
c
o
r
r
e
c
t
l
y
r
e
t
r
i
e
v
e
d
ST
t
o
p
C
r
e
t
r
i
e
v
e
d
(1
7
)
E
is
u
s
ed
to
m
ea
s
u
r
e
t
h
e
p
er
f
o
r
m
a
n
ce
of
t
h
e
r
etr
ie
v
al
r
es
u
l
ts
w
it
h
a
f
ix
ed
le
n
g
t
h
,
e.
g
.
t
h
e
f
ir
s
t
32
m
o
d
el
s
.
It
co
m
b
in
e
s
b
o
th
th
e
P
r
ec
is
io
n
P
an
d
R
ec
all
R
:
r
e
l
e
v
a
n
t
c
o
r
r
e
c
t
l
y
r
e
t
r
i
e
v
e
d
R
e
c
a
l
l
a
l
l
r
e
l
e
v
a
n
t
r
e
l
e
v
a
n
t
c
o
r
r
e
c
t
l
y
r
e
t
r
i
e
v
e
d
P
r
e
c
i
e
s
a
l
l
r
e
t
r
i
e
v
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2252
-
8776
IJ
-
I
C
T
Vo
l.
8
,
No
.
1
,
A
p
r
il
2
0
1
9
:
1
–
12
10
2
11
E
PR
(
1
8
)
DC
G
is
d
e
f
in
ed
as
t
h
e
s
u
m
m
e
d
w
ei
g
h
ted
v
a
lu
e
r
elate
d
to
th
e
p
o
s
itio
n
s
of
t
h
e
r
elev
a
n
t
m
o
d
els.
1
2
2
lo
g
P
k
k
w
D
CG
w
k
(
1
9
)
w
h
er
e
k
w
d
en
o
te
w
ei
g
h
ted
v
al
ue
of
each
r
etr
iev
al
r
es
u
lt
,
a
n
d
k
d
en
o
tes
t
h
e
in
d
e
x
of
r
etr
iev
a
l
r
esu
lt
.
P
is
th
e
n
u
m
b
er
of
r
etr
iev
al
r
es
u
lt.
A
P
ca
n
b
e
co
m
p
u
ted
b
y
co
u
n
t
in
g
t
h
e
to
tal
ar
ea
u
n
d
er
t
h
e
P
r
ec
is
io
n
-
R
ec
a
ll
c
u
r
v
e.
T
h
e
h
ig
h
er
P
r
ec
is
io
n
-
R
ec
all
c
u
r
v
e
w
o
u
ld
g
et
a
b
etter
A
P
v
al
u
e.
W
e
co
m
p
ar
e
o
u
r
ap
p
r
o
ac
h
w
i
th
o
th
er
f
o
u
r
lead
in
g
s
k
etc
h
-
b
ased
3
D
m
o
d
el
r
etr
iev
al
alg
o
r
it
h
m
s
,
w
h
ich
u
til
ize
s
k
eleto
n
c
h
ar
ac
ter
is
tic
s
as
t
h
e
f
ea
tu
r
es
to
d
escr
ib
e
th
e
o
b
j
ec
t
s
h
ap
e.
S
u
n
d
ar
H
et
al.
[
3
1
]
ar
e
th
e
m
o
s
t
r
ep
r
esen
tat
iv
e
in
t
h
e
f
ield
o
f
s
k
eleto
n
b
ased
s
h
ap
e
m
atc
h
in
g
a
n
d
r
etr
ie
v
al.
T
h
e
y
u
tili
ze
d
t
h
e
th
i
n
n
i
n
g
alg
o
r
it
h
m
a
n
d
cl
u
s
ter
i
n
g
al
g
o
r
ith
m
wh
ich
h
e
lp
les
s
en
th
e
ef
f
ec
t
o
f
m
an
y
s
m
a
ll
p
er
tu
r
b
atio
n
s
o
n
t
h
e
s
u
r
f
ac
e
a
n
d
r
ed
u
ce
t
h
e
n
u
m
b
er
o
f
n
o
d
es
n
ec
ess
ar
y
f
o
r
s
k
ele
tal
g
r
ap
h
co
n
s
tr
u
ctio
n
.
L
ei
H
e
t
al.
[
3
2
]
u
s
ed
th
i
n
n
i
n
g
al
g
o
r
ith
m
o
n
th
e
s
il
h
o
u
e
tte
i
m
ag
es
to
ex
tr
ac
t
t
h
e
co
r
r
esp
o
n
d
in
g
s
k
eleto
n
s
an
d
u
tili
ze
d
s
k
ele
to
n
p
r
u
n
in
g
m
et
h
o
d
.
L
in
S
et
al.
[
3
3
]
g
o
t
th
e
s
k
e
leto
n
o
f
3
D
m
o
d
el
th
r
o
u
g
h
s
k
e
leto
n
e
x
tr
ac
tio
n
al
g
o
r
ith
m
b
ased
o
n
m
e
s
h
s
i
m
p
li
f
icatio
n
an
d
m
es
h
co
n
tr
ac
tio
n
.
Sirin
Y
et
al.
[
3
4
]
s
tar
ted
b
y
d
r
a
w
i
n
g
cir
cles
o
f
i
n
cr
ea
s
i
n
g
r
ad
iu
s
ar
o
u
n
d
s
k
elet
o
n
s
,
w
h
ic
h
lead
to
ea
ch
s
k
elet
o
n
co
r
r
esp
o
n
d
ed
to
th
e
ce
n
tr
e
o
f
a
m
ax
i
m
a
ll
y
i
n
s
c
r
ib
ed
cir
cle.
W
e
illu
s
tr
ate
3
D
m
o
d
el
s
o
f
ca
r
,
la
m
p
,
p
la
n
e
a
n
d
ch
air
i
n
t
h
e
F
ig
u
r
e
1
3
an
d
th
e
av
er
ag
e
co
m
p
ar
is
o
n
r
esu
lt is
s
h
o
w
n
in
T
ab
le
1
.
Fig
u
r
e
13.
E
x
a
m
p
les
of
s
k
etc
h
-
b
ased
r
etr
iev
al
r
es
u
lts
T
ab
le
1.
Me
tr
ics
f
o
r
th
e
P
er
f
o
r
m
an
ce
C
o
m
p
ar
is
o
n
b
et
w
ee
n
Ou
r
A
p
p
r
o
ac
h
a
nd
Oth
er
A
p
p
r
o
ac
h
es
A
p
p
r
o
a
c
h
e
s
NN
FT
ST
E
D
C
G
AP
O
u
r
A
p
p
r
o
a
c
h
0
.
3
8
7
0
.
3
1
6
0
.
3
8
3
0
.
3
7
4
0
.
5
8
9
0
.
3
9
4
L
e
i
H’
s
A
p
p
r
o
a
c
h
0
.
3
6
5
0
.
2
6
8
0
.
3
3
1
0
.
3
5
2
0
.
5
4
1
0
.
3
3
5
S
u
n
d
a
r
H’s
A
p
p
r
o
a
c
h
0
.
3
4
7
0
.
2
5
1
0
.
3
2
7
0
.
3
2
1
0
.
5
3
3
0
.
3
2
6
L
i
n
S’s
A
p
p
r
o
a
c
h
0
.
2
8
5
0
.
2
3
6
0
.
2
9
4
0
.
2
6
8
0
.
4
9
1
0
.
2
7
8
S
i
r
i
n
Y’s
A
p
p
r
o
a
c
h
0
.
2
1
1
0
.
2
0
8
0
.
2
4
3
0
.
2
0
2
0
.
4
6
6
0
.
2
0
4
It
is
o
b
v
io
u
s
th
at
o
u
r
ap
p
r
o
ac
h
o
u
tp
er
f
o
r
m
s
o
th
er
lea
d
in
g
s
k
etc
h
-
b
a
s
ed
3D
m
o
d
e
l
r
etr
iev
al
ap
p
r
o
ac
h
es.
T
h
e
ap
p
r
o
ac
h
es
of
S
u
n
d
ar
H
et
al
[
3
1
]
an
d
L
ei
H
et
al
[
3
2
]
u
tili
ze
d
t
h
e
th
in
n
i
n
g
a
lg
o
r
it
h
m
w
h
ic
h
is
q
u
ite
s
e
n
s
iti
v
e
to
n
o
is
e
an
d
can
ac
h
ie
v
e
lar
g
e
a
m
o
u
n
t
of
ca
lcu
latio
n
.
L
i
n
S
et
al
[
3
3
]
u
s
e
d
th
e
f
r
o
n
t
v
ie
w
of
Evaluation Warning : The document was created with Spire.PDF for Python.