T
E
L
KO
M
NI
K
A
,
V
o
l.
1
4
,
N
o.
3
,
S
ept
em
ber
20
1
6
,
pp.
10
77
~
108
2
I
S
S
N
:
1
693
-
6
930
,
ac
c
r
edi
t
ed
A
b
y
D
IK
T
I,
D
e
c
r
e
e
N
o
:
58/
D
I
K
T
I
/
K
ep/
2013
D
O
I
:
10.
12928/
T
E
LK
O
M
N
I
K
A
.
v
1
4
i
3
.
3590
10
77
R
ec
ei
v
ed
A
p
r
il
2
,
2
01
6
;
R
e
v
i
s
ed
J
une
1
2
,
20
1
6
;
A
c
c
e
pt
ed
J
u
ne
2
9
,
201
6
H
y
b
r
id
Hier
ar
ch
ical Co
lli
sio
n
D
et
ect
i
o
n
Base
d
o
n
Da
t
a
Reu
se
Ji
an
ca
i
H
u
,
K
e
j
i
n
g
H
e
*
,
X
i
a
o
b
i
n
L
i
n
,
F
u
n
a
n
L
i
n
S
c
hoo
l
of
C
om
p
ut
er
S
c
i
e
nc
e
a
nd E
ng
i
neer
i
ng
,
S
out
h C
hi
na
U
ni
v
er
s
i
t
y
of
T
ec
hno
l
ogy
,
G
uangz
hou 5106
41,
G
u
angd
o
ng,
C
hi
na
*
C
or
r
es
po
ndi
ng a
ut
hor
,
e
-
m
a
i
l
:
k
j
h
e@
s
c
ut
.
e
du.
c
n
A
b
st
r
act
T
o
i
m
pr
ov
e
t
he
ef
f
i
c
i
e
nc
y
of
c
ol
l
i
s
i
on
d
et
e
c
t
i
o
n
bet
w
een
r
i
gi
d
bo
di
e
s
i
n
c
o
m
p
le
x
s
c
en
es
,
t
his
paper
pr
o
po
s
es
a m
et
h
od ba
s
ed on hy
br
i
d b
ound
i
ng
v
ol
um
e
hi
er
ar
c
hi
e
s
f
or
c
ol
l
i
s
i
on det
ec
t
i
on.
I
n or
der
t
o
i
m
pr
ov
e
t
he
s
i
m
ul
at
i
on
pe
r
f
or
m
anc
e,
t
h
e
m
et
hod
i
s
ba
s
e
d
on
w
ei
gh
t
ed
or
i
ent
ed
bo
und
i
n
g
box
a
nd
m
ak
e
s
dens
e s
am
pl
i
ng on t
h
e c
on
v
e
x
h
ul
l
s
of
t
he ge
om
et
r
i
c
m
od
el
s
.
T
h
e hi
er
ar
c
h
i
c
a
l
boun
di
n
g v
ol
um
e t
r
ee
i
s
c
om
pos
ed o
f
m
any
l
a
y
er
s
.
T
h
e upp
er
m
os
t
l
ay
er
ad
opt
s
a
c
ubi
c
bou
ndi
ng b
ox
,
w
hi
l
e l
ow
e
r
l
ay
er
s
em
pl
o
y
w
ei
ght
e
d or
i
e
nt
ed b
ound
i
ng b
ox
.
I
n t
he m
ea
nt
i
m
e,
t
he d
at
a
of
w
ei
ght
e
d or
i
en
t
ed bo
und
i
n
g box
i
s
r
eus
e
d
f
or
t
r
i
a
ngl
e i
nt
er
s
ec
t
i
on
c
h
ec
k
.
W
e t
e
s
t
t
he m
et
h
od
us
i
ng t
w
o
s
c
e
ne
s
.
T
h
e f
i
r
s
t
s
c
ene
c
on
t
ai
ns
t
w
o B
udd
h
a
m
odel
s
w
i
t
h
t
ot
al
l
y
361,
690
t
r
i
angl
e
f
a
c
et
s
.
T
he
s
ec
o
nd
s
c
e
ne i
s
c
om
po
s
ed
of
20
0 m
ode
l
s
w
i
t
h t
ot
al
l
y
115
,
200 t
r
i
ang
l
e f
ac
e
t
s
.
T
he
ex
p
er
i
m
ent
s
v
er
i
f
y
t
h
e ef
f
ec
t
i
v
e
nes
s
of
t
he
pr
op
os
e
d m
et
ho
d.
Ke
y
w
o
rd
s
:
c
ol
l
i
s
i
on d
et
e
c
t
i
o
n,
hi
er
ar
c
h
i
c
a
l
s
t
r
uc
t
ur
e
,
dat
a
r
e
us
e
C
o
p
y
r
i
g
h
t
©
20
16 U
n
i
ver
si
t
a
s A
h
mad
D
ah
l
an
.
A
l
l
r
i
g
h
t
s r
eser
ved
.
1
.
I
n
tr
o
d
u
c
ti
o
n
C
ol
l
i
s
i
on
d
et
ec
t
i
on i
s
w
i
d
el
y
us
ed i
n
c
om
put
er
g
am
es
,
v
i
r
t
ual
s
ur
g
er
y
,
p
h
y
s
i
c
a
l
s
i
m
ul
at
i
on,
r
o
bot
i
c
s
and s
o on [
1,
2
]
.
I
t
p
l
a
y
s
a
n ex
t
r
em
el
y
i
m
por
t
ant
r
ol
e i
n t
h
es
e f
i
el
ds
.
T
he
pur
pos
es
of
c
ol
l
i
s
i
o
n det
ec
t
i
on
ar
e t
o d
et
ec
t
w
h
et
h
er
t
h
e c
ol
l
i
s
i
o
n be
t
w
ee
n obj
ec
t
s
oc
c
ur
s
or
not
,
and
t
o k
no
w
w
he
n a
nd
w
he
r
e t
he
c
ol
l
i
s
i
on
oc
c
ur
s
.
W
i
t
h
t
he
ad
v
a
nc
e
of
v
i
r
t
ual
r
ea
l
i
t
y
t
ec
hno
l
o
g
y
i
n r
ec
ent
y
e
ar
s
,
t
he d
i
f
f
i
c
ul
t
i
es
of
t
he s
c
ene s
i
m
ul
at
i
o
n al
s
o i
nc
r
eas
es
.
I
n t
he s
a
m
e t
i
m
e,
dat
a
s
t
r
uc
t
ur
es
and al
gor
i
t
hm
s
abou
t
c
ol
l
i
s
i
o
n de
t
ec
t
i
on h
av
e
bec
om
e
m
or
e and m
o
r
e c
o
m
pl
ex
i
n
dea
l
i
n
g
w
i
t
h
s
uc
h
l
ar
ge
d
at
a
s
et
s
,
es
pec
i
a
l
l
y
f
or
r
eal
-
t
i
m
e
c
al
c
ul
at
i
on
.
R
es
ear
c
h
o
n
t
he
c
o
l
l
i
s
i
o
n
det
ec
t
i
on m
et
hod has
l
o
n
g hi
s
t
or
y
,
and m
an
y
r
es
e
ar
c
her
s
hav
e r
es
ear
c
h
ed
deep
l
y
and p
ut
f
or
w
ar
d a
s
er
i
es
of
ef
f
i
c
i
ent
al
gor
i
t
hm
s
bas
ed on bou
ndi
ng box
.
B
o
un
di
n
g box
t
i
ght
l
y
c
ont
a
i
ns
t
he obj
ec
t
b
ut
w
i
t
h s
i
m
pl
e geom
et
r
i
c
c
har
ac
t
er
i
s
t
i
c
s
,
t
o appr
ox
i
m
at
el
y
d
es
c
r
i
be t
he
obj
ec
t
.
B
ef
or
e c
ond
uc
t
i
n
g t
h
e c
ol
l
i
s
i
on
det
ec
t
i
o
n am
ong r
eal
o
bj
ec
t
s
,
t
he i
nt
er
s
ec
t
i
on of
b
ound
i
n
g box
es
i
s
c
hec
k
ed f
i
r
s
t
l
y
.
I
f
t
he
i
nt
e
r
s
ec
t
i
on h
app
ens
,
a
f
ur
t
her
c
ol
l
i
s
i
on
det
ec
t
i
o
n
i
s
nee
de
d.
T
her
e
ar
e
s
om
e
c
o
m
m
on
t
y
p
es
of
boun
di
ng
box
es
s
u
c
h
as
A
x
i
s
-
A
l
i
gne
d
B
ou
nd
i
ng
B
ox
(
AA
BB)
[
3
]
,
O
r
i
ent
ed
B
ou
n
di
n
g
B
ox
(
O
B
B
)
[
4
],
K
-
D
O
Ps
[
5
]
,
a
nd
S
ph
er
es
[
6
].
D
i
ff
e
r
e
n
t
t
y
p
e
s
o
f
boun
di
ng b
ox
es
ha
v
e
di
f
f
er
ent
f
oc
us
,
f
or
ex
a
m
pl
e t
h
e s
i
m
pl
i
c
i
t
y
an
d t
i
ght
n
es
s
.
T
he s
i
m
pl
i
c
i
t
y
and
t
i
g
ht
n
es
s
of
boun
di
ng
box
ar
e
of
t
en
c
o
nt
r
ad
i
c
t
or
y
.
R
ec
en
t
l
y
,
m
os
t
r
es
ear
c
he
s
m
ai
nl
y
f
oc
us
on i
m
pr
ov
i
n
g
t
he
ef
f
i
c
i
enc
y
and
ac
c
ur
ac
y
of
c
ol
l
i
s
i
o
n
det
ec
t
i
on
a
l
g
or
i
t
hm
s
.
F
i
gue
i
r
edo
[
7
]
us
es
t
he o
v
er
l
ap
pi
ng m
ul
t
i
-
ax
i
s
boun
di
ng
box
,
w
hi
c
h c
a
n f
i
l
t
er
d
i
s
j
oi
nt
obj
ec
t
s
f
as
t
an
d i
m
pr
ov
e
t
he
ef
f
i
c
i
enc
y
of
c
ol
l
i
s
i
on de
t
ec
t
i
on.
Mac
i
e
l
[
8
]
em
pl
o
y
s
t
he
s
pher
e f
or
f
as
t
c
ol
l
i
s
i
on d
et
ec
t
i
on
am
ong
obj
ec
t
s
.
Lai
[
9
]
us
es
s
pher
e
and
c
y
l
i
n
der
t
o
r
epr
es
e
n
t
obj
ec
t
s
f
or
qui
c
k
nav
i
g
at
i
o
n
i
n
t
he
s
c
ene,
w
hi
c
h
i
s
f
as
t
but
not
ac
c
ur
at
e
e
nou
gh.
C
ha
ng
[
10
]
c
o
m
bi
nes
t
he
s
ph
er
e
w
i
t
h O
B
B
t
o
i
m
pr
ov
e
t
he s
pee
d of
c
ol
l
i
s
i
on d
et
e
c
t
i
on,
b
ut
bec
a
us
e of
t
he l
i
m
i
t
at
i
ons
of
t
he s
pher
e,
t
h
e ac
c
ur
ac
y
of
det
ec
t
i
on
i
s
no
t
en
oug
h.
S
i
n
gl
e t
y
pe
of
bou
ndi
ng
b
ox
has
s
om
e def
i
c
i
enc
i
es
and
ac
c
ur
ac
y
pr
obl
em
s
i
n r
ea
l
-
t
i
m
e c
ol
l
i
s
i
on
det
ec
t
i
o
n [
11
]
,
t
h
er
ef
or
e m
or
e and
m
or
e r
es
ea
r
c
her
s
pr
opos
e
m
an
y
a
l
gor
i
t
hm
s
c
o
m
bi
ni
ng t
he
ad
v
an
t
ag
es
of
di
f
f
er
ent
bou
ndi
ng b
ox
es
.
A
m
ong t
hem
,
B
ou
nd
i
ng
V
ol
um
e
H
i
er
ar
c
hi
es
(
B
V
H
)
[
12
-
14
]
i
s
one
of
t
he
m
os
t
w
i
de
l
y
us
e
d,
w
hi
c
h
c
an
w
or
k
i
n
c
o
m
pl
ex
en
v
i
r
onm
ent
s
.
A
r
bab
i
[
15
]
us
es
c
y
l
i
ndr
i
c
al
a
nd r
ad
i
a
l
s
pac
e
s
egm
ent
at
i
on m
et
hod
t
o
per
f
or
m
c
ol
l
i
s
i
on
det
ec
t
i
o
n
f
or
j
oi
nt
c
on
nec
t
i
on.
T
hi
s
m
et
hod
has
a
b
et
t
er
pr
ec
i
s
i
on
w
i
t
h
l
i
m
i
t
ed
us
age
i
n
c
ol
l
i
s
i
on
det
ec
t
i
o
n of
bou
nd
ar
y
m
ov
em
ent
.
I
n r
ec
e
nt
y
ear
s
,
m
an
y
r
e
s
ear
c
her
s
us
e
har
d
w
ar
e a
c
c
el
er
at
i
o
n t
o
s
peed u
p t
he c
o
l
l
i
s
i
o
n det
ec
t
i
on
w
i
t
h
t
he h
el
p of
be
t
t
er
c
om
put
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
10
77
–
1
082
1078
per
f
or
m
anc
e
[
16
-
18
].
X
i
e
[
19
]
c
om
bi
nes
hi
er
ar
c
hi
c
a
l
boun
di
ng
s
ph
er
e
w
i
t
h
G
P
U
ac
c
el
er
at
i
on
t
o
s
i
m
ul
at
e c
ol
l
i
s
i
on
de
t
ec
t
i
o
n am
ong r
i
gi
d b
od
i
es
f
or
r
hi
no
pl
as
t
y
.
S
hen
[
2
0]
ado
pt
s
t
he
m
i
x
ed
boun
di
ng
v
o
l
um
e hi
er
ar
c
h
y
t
r
ee
t
o d
et
ec
t
t
he
pot
e
nt
i
al
c
ol
l
i
s
i
on o
bj
ec
t
s
et
s
qu
i
c
k
l
y
,
an
d t
h
en
us
es
t
he s
t
r
eam
i
ng pat
t
er
n al
gor
i
t
hm
f
or
ac
c
ur
at
e c
ol
l
i
s
i
o
n det
ec
t
i
o
n.
B
ut
t
hes
e appr
oac
h
es
hav
e n
ot
t
ak
en t
h
e a
dv
ant
a
ges
of
di
f
f
er
ent
t
y
p
es
of
bo
und
i
n
g box
es
.
I
n t
h
i
s
p
ap
er
,
w
e
pr
o
pos
e
an
i
m
pr
ov
ed
h
y
br
i
d
hi
er
ar
c
hi
c
al
c
ol
l
i
s
i
on
det
ec
t
i
o
n
m
et
hod
bas
ed o
n da
t
a r
eus
e
.
W
e
c
om
bi
ne t
he s
i
m
pl
i
c
i
t
y
of
i
m
pr
ov
e
d A
A
B
B
and t
he t
i
gh
t
nes
s
of
O
B
B
t
o
bui
l
d h
y
br
i
d
B
V
H
s
t
r
uc
t
ur
e f
or
obj
ec
t
s
.
O
ur
i
m
pr
ov
ed h
y
br
i
d hi
er
ar
c
hi
c
a
l
s
t
r
u
c
t
ur
e has
t
w
o
l
a
y
er
s
.
T
he upper
m
os
t
l
a
y
er
us
es
opt
i
m
i
z
ed c
ubi
c
b
ound
i
n
g box
,
w
hi
c
h i
s
eas
y
t
o r
ot
at
e a
nd
t
r
ans
f
or
m
.
B
y
t
h
i
s
w
a
y
,
t
h
e obj
ec
t
s
ar
e not
i
nt
er
s
ec
t
ed i
n t
he s
c
ene c
an b
e ex
c
l
ude
d qu
i
c
k
l
y
.
T
he
l
o
w
er
l
a
y
er
s
us
e
t
he
O
B
B
,
w
h
i
c
h
has
bet
t
er
t
i
g
ht
n
es
s
and
c
an
be
us
ed
f
or
f
ur
t
her
c
ol
l
i
s
i
on
det
ec
t
i
on f
or
t
hos
e i
nt
er
s
e
c
t
ed
obj
ec
t
s
f
ound i
n t
he
u
pper
l
a
y
er
.
T
he m
et
hod c
a
n i
m
pr
ov
e t
he
ef
f
i
c
i
enc
y
an
d ac
c
ur
ac
y
of
c
ol
l
i
s
i
on
det
ec
t
i
o
n b
y
t
ak
i
ng
t
he a
dv
ant
ages
of
t
he
h
y
br
i
d h
i
er
ar
c
h
i
c
al
boun
di
ng s
t
r
uc
t
ur
es
.
I
n t
h
e
m
eant
i
m
e,
w
e i
nt
r
od
uc
e
a new
a
l
g
or
i
t
hm
f
or
t
r
i
angl
e i
nt
er
s
ec
t
i
on
c
hec
k
b
y
r
eus
i
n
g t
h
e O
B
B
dat
a
[
21
]
,
w
h
i
c
h c
an ef
f
ec
t
i
v
e
l
y
r
educ
e t
he c
al
c
u
l
at
i
on an
d f
ur
t
her
i
m
pr
ov
e t
he
ef
f
i
c
i
enc
y
.
T
hi
s
paper
i
s
or
g
an
i
z
ed as
f
ol
l
o
w
s
.
F
o
l
l
o
w
i
ng t
h
e i
nt
r
o
duc
t
i
o
n i
n
S
ec
.
1,
S
ec
.
2 pr
es
ent
s
t
he
i
m
pr
ov
ed
h
y
br
i
d
hi
er
ar
c
hi
c
al
c
o
l
l
i
s
i
on
d
et
ec
t
i
on
m
et
hod
bas
e
d
on
da
t
a
r
eus
e.
I
n
S
ec
.
3,
t
he
m
et
hod i
s
ap
pl
i
ed
t
o
t
he
c
o
l
l
i
s
i
on
de
t
ec
t
i
on
of
t
w
o m
odel
s
a
nd m
ul
t
i
p
l
e
m
odel
s
r
es
pec
t
i
v
e
l
y
.
W
e
al
s
o c
om
par
e t
h
e
per
f
or
m
anc
e of
our
m
et
hod
w
i
t
h
ot
h
er
ap
pr
oac
hes
.
F
i
n
al
l
y
,
s
o
m
e c
onc
l
us
i
ons
and
di
s
c
us
s
i
ons
ar
e m
ade i
n S
ec
.
4.
2.
R
e
sea
r
ch
M
et
h
o
d
T
he c
ol
l
i
s
i
on d
et
ec
t
i
on am
ong
A
A
B
B
i
s
s
i
m
pl
e and f
as
t
,
w
hi
c
h c
a
n be ac
c
om
pl
i
s
he
d
w
i
t
h
i
n s
i
x
t
es
t
s
.
W
hen t
he
obj
ec
t
m
ov
es
,
A
A
B
B
has
t
o
be
r
ebu
i
l
t
.
A
g
i
v
en
obj
ec
t
’
s
O
B
B
i
s
t
he
s
m
al
l
es
t
c
ub
oi
d
t
h
at
c
on
t
ai
ns
t
he
obj
ec
t
.
C
om
par
ed t
o
A
A
B
B
,
O
B
B
has
b
et
t
er
t
i
g
ht
nes
s
,
b
ut
t
he
i
nt
er
s
ec
t
i
on c
h
ec
k
i
s
m
or
e ex
pens
i
v
e.
T
he
m
os
t
us
ed
m
et
hod
t
o
det
er
m
i
ne
w
h
et
h
er
t
w
o
c
on
v
ex
h
ul
l
s
ar
e
i
n
t
er
s
ec
t
ed
or
not
i
s
t
o us
e t
he
s
ep
ar
at
i
ng ax
i
s
t
es
t
[
22
]
.
T
he num
ber
of
s
epar
at
i
ng ax
i
s
d
epe
nds
o
n t
he t
y
p
e of
boun
di
ng b
ox
.
F
or
ex
am
pl
e,
A
A
B
B
has
3 (
x
,
y
,
z
)
t
y
pi
c
a
l
s
epar
at
i
n
g ax
es
.
O
B
B
has
1
5 t
y
p
i
c
a
l
s
epar
at
i
ng
ax
es
,
s
o
t
h
e
i
nt
er
s
ec
t
i
o
n c
hec
k
of
t
w
o
O
B
B
s
c
ou
l
d
t
ak
e up
t
o
1
5 c
om
par
i
s
on
oper
at
i
o
ns
,
60
ad
di
t
i
o
n
oper
at
i
o
ns
,
81
m
ul
t
i
p
l
i
c
at
i
on o
per
at
i
o
ns
,
an
d 2
4
abs
ol
ut
e
v
a
l
u
e
oper
at
i
o
ns
[
23
].
T
r
adi
t
i
on
al
h
y
br
i
d
B
V
H
j
us
t
s
i
m
pl
y
us
es
t
w
o
d
i
f
f
er
ent
boun
di
ng
b
o
x
es
t
o
enc
l
os
e
ev
er
y
obj
ec
t
f
or
bet
t
er
c
om
pac
t
nes
s
and f
as
t
ex
c
l
udi
ng d
i
s
j
oi
nt
obj
ec
t
s
[
24
]
.
W
hen t
he obj
ec
t
s
ar
e
c
l
os
e
but
not
c
ol
l
i
de,
t
he
y
c
an
be
s
ep
ar
at
ed
s
i
m
pl
y
b
ec
aus
e
of
boun
di
ng
b
ox
’
s
t
i
ght
nes
s
,
t
hus
t
he
p
er
f
or
m
anc
e
of
h
y
br
i
d
B
V
H
i
s
be
t
t
er
t
h
an
t
ha
t
of
s
i
ngl
e
b
oun
di
n
g
b
ox
.
H
o
w
e
v
er
,
w
hen
t
w
o
obj
ec
t
s
ar
e
i
nt
er
s
ec
t
ed,
t
h
e i
nt
er
s
ec
t
i
on
of
bou
nd
i
ng
box
s
t
r
uc
t
ur
es
i
s
goi
ng
t
o b
e c
hec
k
ed
m
ul
t
i
pl
y
t
i
m
es
,
w
h
i
c
h
br
i
ng
s
r
edund
ant
c
a
l
c
ul
at
i
ons
.
T
hus
i
t
is
un
abl
e
t
o
f
ul
f
i
l
l
t
he
r
equi
r
em
ent
s
of
r
eal
-
t
im
e
c
o
llis
io
n de
t
e
c
t
i
on.
T
hi
s
paper
pr
o
v
i
d
es
an ef
f
i
c
i
ent
h
y
br
i
d hi
er
ar
c
hi
c
al
c
ol
l
i
s
i
on
det
ec
t
i
on m
et
hod bas
ed o
n
dat
a r
eus
e.
W
e
i
nt
egr
at
e t
he c
ubi
c
bou
nd
i
ng b
ox
and
w
ei
g
ht
e
d O
B
B
t
o f
or
m
a h
y
br
i
d
hi
er
ar
c
hi
c
a
l
s
t
r
uc
t
ur
e
i
n
t
he
pr
et
r
e
at
m
ent
p
has
e.
T
hen
w
e
i
nt
r
odu
c
e t
he
t
r
i
a
ngl
e
i
nt
er
s
ec
t
i
on
c
hec
k
al
gor
i
t
h
m
b
y
r
eus
i
ng
t
h
e O
B
B
dat
a,
w
hi
c
h
c
an
ef
f
ec
t
i
v
e
l
y
r
e
duc
e
t
h
e
c
al
c
ul
a
t
i
o
n t
i
m
e and i
m
pr
ov
e t
he
per
f
or
m
anc
e of
c
ol
l
i
s
i
on d
et
ec
t
i
on.
2
.1
.
F
r
o
m
A
A
B
B
to
C
u
b
i
c
B
o
u
n
d
i
n
g
B
o
x
T
o r
educ
e t
he m
e
m
or
y
c
o
ns
um
pt
i
on an
d t
o s
p
eed
u
p t
he
c
a
l
c
ul
at
i
on,
w
e
us
e c
ubi
c
boun
di
ng
box
,
w
hi
c
h
i
s
a s
pec
i
a
l
k
i
nd of
A
A
B
B
.
T
he c
ubi
c
b
ou
ndi
ng
box
h
as
t
he
s
a
m
e hal
f
-
w
i
dt
h
ex
t
ent
(
or
r
adi
us
r
)
i
n t
hr
ee ax
es
,
and t
he c
ent
er
-
r
ad
i
us
r
epr
es
ent
at
i
on i
s
ad
opt
e
d.
W
e
s
uppos
e
t
hat
t
he
m
odel
i
s
c
om
pos
ed
of
N
t
r
i
an
gl
es
,
and
i
o
,
i
p
,
i
q
ar
e
t
he
v
er
t
ex
es
of
t
r
i
a
ng
l
e
i
,
and
t
he c
en
t
er
po
i
n
t
of
c
ubi
c
b
o
und
i
ng
box
i
s
:
1
1
()
3
=
=
++
∑
N
ii
i
i
C
op
q
N
.
(
1)
T
he hal
f
-
w
i
dt
h
ex
t
en
t
or
r
ad
i
us
i
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
H
y
br
i
d H
i
er
ar
c
hi
c
a
l
C
ol
l
i
s
i
o
n D
et
ec
t
i
o
n B
as
ed
on D
at
a
R
eus
e (
J
ia
n
c
a
i H
u
)
1079
(
)
(
)
m
a
x
m
a
x
(
,)
,
(
,)
,
(
,)
=
i
ii
i
r
di
s
t
o
C
di
s
t
p
C
di
s
t
q
C
,
(
2)
W
h
er
e
di
s
t
(
⋅
,
⋅
)
i
s
t
he f
unc
t
i
o
n t
o c
al
c
ul
at
e t
h
e
di
s
t
a
nc
e b
et
w
ee
n t
w
o po
i
nt
s
.
S
i
nc
e t
he c
ubi
c
boun
di
ng b
ox
has
t
he s
a
m
e
r
adi
us
i
n t
hr
ee ax
es
,
t
her
ef
or
e,
t
he
obj
ec
t
c
an
m
ov
e or
r
ot
at
e
w
i
t
h
out
c
h
ang
i
ng
t
he
r
adi
us
.
T
hi
s
m
et
hod
n
ot
on
l
y
r
educ
es
t
he
s
t
or
ag
e
us
age,
but
al
s
o
ac
c
el
er
at
es
t
he
bou
ndi
ng
b
ox
upda
t
i
n
g.
T
he i
nt
er
s
ec
t
i
on c
hec
k
of
c
ubi
c
bound
i
n
g
box
i
s
s
i
m
pl
er
t
han
s
pher
e
’
s
.
S
up
pos
e
t
h
at
(
x
,
y
,
z
)
i
s
t
he
c
ent
er
p
oi
n
t
of
c
ubi
c
bou
nd
i
ng
box
,
t
hen
t
he
m
i
ni
m
u
m
c
oor
di
nat
e
(
−
xr
,
−
yr
,
−
zr
)
an
d
t
he
m
ax
i
m
u
m
c
oor
di
nat
e
(
+
xr
,
+
y
r
,
+
zr
)
c
an
be
got
t
e
n q
ui
c
k
l
y
.
I
t
c
a
n ac
h
i
e
v
e
i
nt
er
s
ec
t
i
on
c
hec
k
w
i
t
hi
n
s
i
x
t
es
t
s
.
I
n or
der
t
o
i
m
pr
ov
e t
he ef
f
i
c
i
enc
y
,
t
he s
p
at
i
al
a
nd t
em
por
al
c
or
r
el
at
i
on i
s
i
nt
r
od
uc
e
d.
W
e
us
e t
hr
ee l
i
s
t
s
t
o s
av
e t
he pr
oj
ec
t
i
v
e c
oor
di
nat
es
of
obj
ec
t
s
i
n t
hr
ee
(
x
,
y
,
z
)
a
x
es
r
es
pec
t
i
v
el
y
.
W
hen t
he obj
ec
t
’
s
s
t
at
e c
han
ges
,
t
h
e
m
et
ho
d upd
at
es
t
h
e t
hr
ee
l
i
s
t
s
and
qu
i
c
k
l
y
f
i
nds
t
he
bou
nd
i
ng
box
pa
i
r
s
t
hat
i
nt
er
s
ec
t
i
n
t
he
pr
oj
ec
t
i
on
,
and
p
ut
s
t
hos
e
p
ai
r
s
i
nt
o
a
g
l
ob
al
di
c
t
i
onar
y
.
I
t
w
i
l
l
get
obj
ec
t
pai
r
s
o
ut
f
r
om
t
he gl
ob
al
d
i
c
t
i
onar
y
w
hen f
ur
t
h
er
i
n
t
er
s
ec
t
i
on
c
hec
k
i
s
r
equi
r
e
d.
A
t
r
av
er
s
al
a
l
gor
i
t
hm
i
s
us
ed
f
or
det
ec
t
i
n
g
t
h
e
c
ol
l
i
s
i
o
ns
bet
w
een
s
i
bl
i
ng
s
ubt
r
ees
.
W
e
bui
l
d
a ne
w
l
i
s
t
f
or
eac
h
obj
ec
t
t
o r
ec
or
d
t
he
ne
i
g
hbor
s
t
hos
e
c
ol
l
i
d
e
w
i
t
h t
he o
bj
ec
t
.
W
e
dedu
pl
i
c
at
e t
h
e l
i
s
t
t
o ex
c
l
ude r
ed
und
ant
c
ol
l
i
s
i
ons
.
W
h
en t
w
o obj
ec
t
s
c
ol
l
i
d
e,
t
he obj
ec
t
w
i
t
h
l
ar
ger
I
D
num
ber
w
i
l
l
be s
t
or
ed
i
n t
he
l
i
s
t
of
t
he o
bj
ec
t
w
i
t
h s
m
al
l
er
I
D
n
um
ber
.
I
f
t
he l
i
s
t
i
s
em
pt
y
,
t
h
er
e w
as
no c
ol
l
i
s
i
on w
i
t
h t
he obj
ec
t
.
A
c
c
or
d
i
n
g t
o t
he ab
ov
e ana
l
y
s
i
s
,
our
m
et
hod t
ak
es
adv
ant
ages
of
t
he s
pat
i
al
and t
em
por
al
c
or
r
el
at
i
on o
f
obj
ec
t
’
s
m
ov
e
m
ent
,
and
m
a
k
e
s
i
t
not
nec
es
s
ar
y
t
o t
r
a
v
er
s
e f
r
o
m
t
he t
r
ee r
oot
,
a
v
o
i
d
i
n
g us
el
es
s
c
al
c
ul
at
i
on,
s
o s
pee
ds
up t
he c
ol
l
i
s
i
on
det
ec
t
i
on pr
oc
es
s
.
2
.
2
.
W
e
i
g
h
te
d
O
r
i
e
n
te
d
B
o
u
n
d
i
n
g
B
o
x
T
r
adi
t
i
on
al
m
et
hod f
or
c
om
put
i
ng
t
he
c
ent
er
p
oi
n
t
of
O
B
B
i
s
t
o
get
t
he
m
ean pos
i
t
i
on
of
al
l
v
er
t
ex
es
.
B
ut
i
n
pr
ac
t
i
c
e
,
t
he
s
i
z
es
of
t
he
t
r
i
ang
l
es
t
hat
c
ons
t
i
t
ut
e
t
he
obj
ec
t
ar
e
non
un
i
f
or
m
,
s
o us
i
ng t
he
t
r
ad
i
t
i
on
al
m
et
hod
w
i
l
l
m
a
k
e t
he c
al
c
ul
a
t
ed
c
ent
er
p
oi
nt
t
e
nd t
o c
r
o
w
ded t
r
i
an
gl
e
f
ac
et
s
.
T
hi
s
paper
pr
op
os
e
s
w
e
i
g
ht
ed
or
i
ent
e
d
b
oun
di
ng b
ox
.
W
e
m
a
k
e
a
dens
e
s
am
pl
i
ng
f
or
t
he p
oi
nt
s
of
c
on
v
ex
h
ul
l
s
u
r
f
ac
es
i
n or
der
t
o r
ed
uc
e t
h
e i
m
pac
t
of
c
r
ow
d
ed
t
r
i
an
gl
e f
ac
et
s
.
I
n t
he
i
-
t
h t
r
i
a
ngu
l
a
r
f
ac
et
of
c
on
v
e
x
hul
l
,
t
h
e c
ent
er
po
i
nt
i
s
:
3
++
=
i
ii
i
o
pq
c
,
(
3)
A
nd
t
he s
ur
f
ac
e ar
e
a i
s
:
(
)(
)
2
−
×
−
=
i
i
ii
i
o
p
oq
S
.
(
4)
T
he t
ot
al
ar
ea
of
t
he c
on
v
e
x
hul
l
i
s
:
1
=
=
∑
n
i
i
WS
.
(
5)
T
he c
ent
er
of
t
he
bou
ndi
ng
box
i
s
:
1
1
(
)
=
=
⋅
∑
n
i
i
i
C
Sc
n
.
(
6)
Let
j
and
k
be t
he c
om
ponen
t
of
(
x
,
y
,
z
)
,
and
ac
c
or
di
n
g t
o
t
he a
bo
v
e a
na
l
y
s
i
s
,
t
he
c
ov
ar
i
anc
e
,
jk
C
ov
is
:
(
)
,
,,
,
,
,
,
,
,
1
9
12
=
=
+
+
+
−
∑
n
i
j
k
i
j
ik
i
j
ik
i
j
ik
i
j
ik
j
k
i
S
C
ov
c
c
o
o
p
p
q
q
C
C
W
.
(
7)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
10
77
–
1
082
1080
2
.
3
.
H
y
b
r
i
d
H
i
e
r
a
r
c
h
i
c
a
l
B
o
u
n
d
i
n
g
V
o
l
u
m
e
T
r
e
e
I
n
t
h
i
s
pap
er
,
w
e
c
r
eat
e
h
y
br
i
d
b
oun
di
ng
b
ox
t
r
ee
f
or
obj
ec
t
s
i
n
t
h
e
v
i
r
t
u
al
e
nv
i
r
o
nm
ent
.
T
he t
r
ee i
s
c
o
m
pos
ed of
s
e
v
er
a
l
l
a
y
er
s
.
T
he upper
m
os
t
l
a
y
er
us
es
t
he c
ub
i
c
bou
n
di
n
g box
(
S
ec
.
2.
1)
,
an
d t
he l
o
w
er
l
a
y
er
s
u
s
e w
e
i
g
ht
ed O
B
B
(
S
ec
.
2.
2
)
w
i
t
h bet
t
er
t
i
g
ht
n
es
s
.
T
hi
s
al
gor
i
t
hm
not
onl
y
c
a
n
ex
c
l
ude
n
on
i
nt
er
s
ec
t
i
ng
o
bj
ec
t
s
i
n
c
om
pl
ex
s
c
enes
qu
i
c
k
l
y
,
b
ut
al
s
o
c
an
do
ac
c
ur
at
e
c
ol
l
i
s
i
on
de
t
ec
t
i
on.
I
t
has
b
et
t
er
r
ea
l
-
t
im
e
det
ec
t
i
o
n
ef
f
i
c
i
enc
y
a
nd
t
i
g
ht
nes
s
.
E
v
e
n i
n
t
he
w
or
s
t
c
as
e,
t
her
e
i
s
on
l
y
on
e i
nt
er
s
ec
t
i
on c
h
ec
k
f
or
ev
er
y
boun
di
ng b
ox
i
n
eac
h
l
a
y
er
,
w
hi
c
h
f
ul
l
y
r
ef
l
ec
t
s
t
he ad
v
a
nt
a
ges
of
t
he h
y
br
i
d b
oun
di
ng b
ox
s
t
r
uc
t
ur
e.
C
om
par
ed w
i
t
h R
A
P
I
D
a
l
g
or
i
t
hm
[
4]
,
w
h
i
c
h i
s
a w
i
de
l
y
-
us
ed bo
und
i
n
g v
o
l
um
e hi
er
ar
c
h
y
t
r
ee i
m
pl
em
ent
at
i
on b
as
ed
on O
B
B
,
t
he
i
m
pr
ov
ed h
y
br
i
d
B
V
H
has
s
e
v
er
a
l
a
dv
ant
ages
:
1.
R
educ
e
t
he
am
ount
of
m
em
or
y
us
ag
e.
B
ec
a
us
e c
ubi
c
b
ou
ndi
ng
box
r
eq
ui
r
e
s
4
f
l
oat
s
,
an
d O
B
B
o
nl
y
nee
d
s
15
f
l
oat
s
,
t
he
am
ount
of
m
e
m
o
r
y
t
hat
us
ed
t
o s
t
or
e
t
he
bou
nd
i
ng
box
es
ar
e r
e
duc
ed.
2.
S
i
m
pl
e
i
nt
er
s
ec
t
i
on
c
hec
k
.
I
n
t
h
i
s
a
l
gor
i
t
hm
,
t
he
r
o
ot
n
ode
us
es
c
ub
i
c
b
o
und
i
n
g
b
ox
i
ns
t
ea
d
of
O
B
B
.
T
he
i
nt
er
s
ec
t
i
on
c
h
ec
k
s
a
m
ong
c
ubi
c
bound
i
ng
box
es
nee
d
f
ew
er
c
al
c
ul
a
t
i
on
oper
at
i
o
ns
and
ar
e s
i
m
pl
er
.
3.
R
educ
e c
o
l
l
i
s
i
o
n det
ec
t
i
o
n
t
i
m
e.
I
n r
eal
-
t
i
m
e det
ec
t
i
on
pr
oc
es
s
,
t
he c
ubi
c
b
oun
di
ng
box
do
es
not
ne
ed
upd
at
es
,
and
c
an
qui
c
k
l
y
ex
c
l
ud
e
m
o
s
t
noni
nt
er
s
ec
t
i
ng
obj
ec
t
s
,
t
hus
s
a
v
e
a
l
ot
of
unnec
es
s
ar
y
i
n
t
er
s
ec
t
c
hec
k
i
ng t
i
m
e.
2
.
4
. T
r
ia
n
g
le
I
n
te
r
s
e
c
ti
o
n
C
h
eck b
y
R
e
u
s
i
n
g
th
e
O
B
B
D
at
a
I
f
t
w
o bou
nd
i
ng b
ox
es
ar
e i
nt
er
s
ec
t
i
ng,
i
t
does
n
’
t
m
ean t
hat
t
he t
w
o obj
ec
t
s
r
eal
l
y
c
ol
l
i
de,
s
o a
f
ur
t
her
i
nt
er
s
ec
t
i
on
c
hec
k
i
s
r
equi
r
ed.
W
e us
e t
h
e i
nt
er
v
al
i
nt
er
s
ec
t
i
on
al
gor
i
t
hm
[
25
]
t
o c
hec
k
w
he
t
her
t
r
i
an
gl
e f
ac
et
s
i
nt
er
s
ec
t
or
not
.
T
he d
et
ai
l
of
t
h
i
s
al
gor
i
t
hm
i
s
as
f
ol
l
o
w
:
1.
C
al
c
u
l
at
e
p
l
a
ne
eq
uat
i
ons
of
t
w
o
t
r
i
an
gl
es
r
es
pec
t
i
v
el
y
.
I
f
al
l
v
er
t
i
c
es
of
one
t
r
i
angl
e
r
es
i
de
i
n t
he s
am
e s
i
de
of
anot
h
er
t
r
i
a
ngl
e,
t
he t
w
o t
r
i
a
ngl
es
do
not
i
nt
er
s
ec
t
.
2.
I
f
t
hes
e
t
w
o
pl
an
es
i
nt
er
s
e
c
t
,
f
i
gur
e
out
t
he
i
nt
er
s
ec
t
i
ng
l
i
ne
L
of
t
w
o
p
l
a
nes
.
T
hen
es
t
abl
i
s
h a
c
oor
d
i
nat
e s
y
s
t
em
w
hi
c
h
has
an
ax
i
s
p
ar
al
l
el
i
n
g t
o
L
.
3.
C
al
c
u
l
at
e
t
he
pr
oj
ec
t
i
on
i
nt
er
v
a
l
s
of
t
he t
w
o t
r
i
ang
l
es
i
n t
he
c
oor
di
nat
e s
y
s
t
em
.
4.
I
f
t
hes
e t
w
o
i
nt
er
v
a
l
s
o
v
er
l
a
p,
t
h
e t
w
o
t
r
i
a
ngl
es
i
nt
er
s
e
c
t.
I
n or
der
t
o
i
m
pr
ov
e t
h
e e
f
f
i
c
i
enc
y
of
t
he a
bo
v
e
al
g
or
i
t
hm
,
t
he O
B
B
-
bas
e
d t
r
i
ang
l
e
i
nt
er
s
ec
t
i
on
c
hec
k
m
et
hod
[
21
]
i
s
a
dop
t
ed.
I
n
t
r
ad
i
t
i
on
al
t
r
i
ang
l
e
i
nt
er
s
ec
t
i
on
c
he
c
k
,
c
oor
di
nat
e
t
r
ans
f
or
m
at
i
on i
s
nec
es
s
a
r
y
,
w
h
i
c
h m
eans
t
hat
o
ne b
oun
di
ng b
ox
or
t
r
i
a
ngl
e m
us
t
be
r
epr
es
ent
e
d i
n a
not
her
on
e’
s
c
oor
di
n
at
e s
y
s
t
em
.
I
n our
m
et
hod,
eac
h
l
e
af
no
de i
n t
h
e h
y
br
i
d
B
V
H
i
s
a
r
e
c
t
angl
e,
w
hi
c
h c
o
nt
a
i
ns
o
nl
y
o
ne
t
r
i
ang
l
e.
A
t
l
eaf
nod
es
,
w
e
r
epr
es
ent
eac
h
t
r
i
an
gl
e
us
i
ng
t
h
e
c
oor
d
i
n
at
e
s
y
s
t
em
of
i
t
s
bou
nd
i
ng
box
.
B
y
r
eus
i
n
g t
he c
o
or
d
i
nat
e s
y
s
t
em
i
nf
or
m
at
i
on,
i
t
i
s
e
as
y
t
o c
a
l
c
ul
at
e
t
he
poi
nt
-
to
-
pl
a
ne
di
s
t
anc
e
.
I
n t
hi
s
w
a
y
,
t
h
e r
edun
dant
c
a
l
c
ul
a
t
i
o
ns
i
n t
r
i
ang
l
e i
nt
er
s
ec
t
i
on c
hec
k
ar
e r
educ
ed.
T
her
eb
y
t
h
e ef
f
i
c
i
enc
y
of
c
ol
l
i
s
i
on
det
ec
t
i
o
n i
s
i
nc
r
eas
e
d.
3
.
E
xp
er
i
m
en
t
s
a
n
d
R
esu
l
t
s
T
o t
es
t
t
he per
f
or
m
anc
e
of
t
he i
m
pr
ov
ed m
et
hod,
i
t
i
s
c
om
par
ed w
i
t
h R
A
P
I
D
.
T
he
al
g
or
i
t
hm
I
i
s
bas
e
d o
n t
he
i
m
pr
ov
em
ent
S
ec
.
2.
1
t
o
2.
3.
A
nd
i
m
pr
ov
ed
a
l
gor
i
t
hm
I
I
t
ak
es
al
l
t
he
i
m
pr
ov
em
ent
s
i
nt
o ac
c
ou
nt
.
T
hi
s
al
gor
i
t
hm
i
s
i
m
pl
e
m
ent
ed
us
i
ng
O
p
enG
L
gr
ap
hi
c
s
l
i
br
ar
y
an
d
i
s
r
u
n
on
a
m
ac
hi
n
e
w
i
t
h I
nt
el
C
or
e2 D
uo T
57
50 pr
oc
es
s
or
,
2 G
B
D
D
R
2 667M
H
z
m
e
m
or
y
and N
V
I
D
I
A
G
eF
or
c
e
8400
M G
S
gr
aph
i
c
s
c
ar
d.
W
e
t
es
t
t
he
m
et
hod us
i
n
g t
w
o s
c
enes
.
T
he
f
i
r
s
t
s
c
ene (
F
i
g
ur
e
1
)
c
ont
ai
ns
t
w
o B
ud
dhas
,
ea
c
h of
w
hi
c
h has
180,
845 t
r
i
ang
ul
ar
f
ac
et
s
and
m
ov
es
i
n a s
pec
i
f
i
c
t
r
aj
ec
t
or
y
.
T
he s
ec
on
d s
c
ene (
F
i
g
ur
e
1)
i
nc
l
udes
200
m
odel
s
w
i
t
h t
ot
a
l
l
y
1
15,
200
t
r
i
an
gl
e
f
ac
et
s
.
T
he
ex
per
i
m
ent
al
r
es
ul
t
s
a
r
e
s
ho
w
n
i
n
F
i
g
ur
e
2.
T
he
ex
per
i
m
ent
s
s
how
t
hat
w
h
en
t
he
num
ber
of
i
nt
er
s
ec
t
ed
t
r
i
an
gl
e gr
o
w
s
,
t
h
e
c
ol
l
i
s
i
on d
et
ec
t
i
on
t
i
m
e
f
or
al
l
m
et
hods
i
nc
r
eas
es
.
I
n
bot
h
t
he
t
w
o
-
m
odel
s
c
ene
and
t
he
m
ul
t
i
-
m
odel
s
c
ene,
our
i
m
pr
ov
ed
h
y
b
r
i
d
h
i
er
ar
c
hi
c
al
c
o
l
l
i
s
i
on
det
ec
t
i
on m
et
hod
per
f
or
m
s
bet
t
er
t
han
R
A
P
I
D
.
W
h
en t
h
e num
ber
of
m
odel
s
i
s
s
m
al
l
,
t
he
di
f
f
er
enc
e bet
w
een
a
l
gor
i
t
hm
I
and R
A
P
I
D
i
s
s
l
i
g
ht
.
B
ut
w
h
en t
her
e
ar
e
a l
ot
of
m
odel
s
,
t
h
e
adv
ant
age
of
i
m
pr
ov
ed
A
l
g
or
i
t
hm
I
gr
adual
l
y
app
ear
s
.
T
hi
s
i
s
bec
aus
e
t
he
i
nt
er
s
e
c
t
i
on
c
hec
k
of
i
m
pr
ov
ed c
ubi
c
bou
nd
i
ng
box
i
s
f
as
t
er
t
han O
B
B
.
A
nd t
her
e i
s
no ne
ed t
o u
p
dat
e
w
h
en t
he
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KO
M
NI
K
A
I
S
S
N
:
1
693
-
6
930
H
y
br
i
d H
i
er
ar
c
hi
c
a
l
C
ol
l
i
s
i
o
n D
et
ec
t
i
o
n B
as
ed
on D
at
a
R
eus
e (
J
ia
n
c
a
i H
u
)
1081
obj
ec
t
’
s
s
t
at
e c
han
ges
.
R
e
us
i
ng
t
he
O
B
B
d
at
a f
or
t
r
i
ang
l
e
i
nt
er
s
ec
t
i
on
c
hec
k
and i
nt
r
od
uc
i
n
g
s
pat
i
o
-
t
em
por
al
c
or
r
el
a
t
i
o
n
c
an r
educ
e r
ed
und
ant
c
al
c
ul
at
i
o
n a
nd
ha
v
e s
i
gn
i
f
i
c
ant
c
ont
r
i
b
ut
i
on t
o
per
f
or
m
anc
e i
m
pr
ov
em
ent
.
F
i
gur
e
1.
T
he s
c
enes
of
B
u
ddhas
(
l
ef
t
)
and t
eap
ot
s
(
r
i
g
ht
)
.
F
i
gur
e
2
.
T
he ex
p
er
i
m
ent
al
r
es
ul
t
s
of
B
u
ddh
as
(
l
ef
t
)
an
d t
ea
pot
s
(
r
i
g
ht
)
.
4
.
C
o
n
c
l
u
s
i
o
n
T
hi
s
paper
has
pr
es
en
t
ed
a
n
i
m
pr
ov
ed
h
y
br
i
d h
i
er
ar
c
hi
c
al
c
ol
l
i
s
i
on
det
ec
t
i
o
n m
et
hod
t
hat
t
ak
es
t
he
adv
ant
ages
of
i
m
pr
ov
ed
c
ubi
c
bo
und
i
n
g
box
,
w
ei
ght
e
d
O
B
B
,
h
y
br
i
d
hi
er
ar
c
hi
c
a
l
boun
di
ng
v
o
l
um
e
t
r
ee,
an
d
da
t
a
r
e
us
i
n
g.
T
he
m
et
hod
i
m
pr
ov
e
s
t
h
e
ef
f
i
c
i
enc
y
of
c
ol
l
i
s
i
on
det
ec
t
i
on.
C
om
par
ed
w
i
t
h
t
r
adi
t
i
on
al
c
o
l
l
i
s
i
o
n de
t
ec
t
i
on
al
g
or
i
t
hm
s
,
t
he ex
per
i
m
ent
al
r
es
ul
t
s
hav
e
s
ho
w
n
t
h
at
t
hi
s
m
et
hod
has
bet
t
er
per
f
or
m
anc
e
i
n
bot
h
t
w
o
-
m
odel
s
c
ene
a
nd
m
ul
t
i
-
m
odel
s
c
ene.
A
c
k
n
o
w
l
e
d
g
e
m
e
n
ts
T
hi
s
w
or
k
w
as
s
uppor
t
ed
b
y
t
he
N
a
t
i
o
na
l
N
at
ur
a
l
S
c
i
e
nc
e
F
ound
at
i
on
of
C
hi
n
a
(
N
S
F
C
)
(
N
o.
61
272
200
,
1
080
501
9)
,
t
he
P
r
o
gr
am
f
or
E
x
c
el
l
e
nt
Y
o
ung
T
eac
her
s
i
n
H
i
g
her
E
d
uc
at
i
on
of
G
uang
don
g,
C
hi
n
a (
N
o.
Y
q2
013
012)
,
t
he
F
un
dam
ent
a
l
R
es
e
ar
c
h F
un
ds
f
or
t
he C
ent
r
a
l
U
ni
v
er
s
i
t
i
es
(
D
215
344
0)
,
t
he
S
p
ec
i
a
l
S
u
ppor
t
P
r
ogr
a
m
of
G
uangdo
ng
P
r
o
v
i
nc
e,
and
t
he
P
ear
l
R
i
v
er
S
c
i
enc
e
&
T
ec
hno
l
og
y
S
t
ar
P
r
oj
ec
t
.
R
ef
er
en
ces
[1
]
J
i
m
enez
P
,
T
hom
as
F
,
T
o
rra
s
C
.
C
ol
l
i
s
i
on det
ec
t
i
on:
a s
ur
v
ey
.
C
om
put
er
s
and G
r
aphi
c
s
.
2
001
;
25(
2
)
:
269
-
2
85
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SSN
:
1
6
9
3
-
6
930
T
E
L
KO
M
NI
K
A
V
o
l.
1
4
,
N
o
.
3
,
S
ept
em
ber
201
6
:
10
77
–
1
082
1082
[2
]
S
uai
b
N
M
,
B
ade
A
,
M
oham
a
d
D
.
H
y
br
i
d
C
ol
l
i
s
i
o
n
C
ul
l
i
ng
by
B
oundi
ng
V
o
l
u
m
es
M
ani
pul
at
i
on
i
n
M
as
s
i
v
e
R
i
gi
d
B
ody
S
i
m
ul
at
i
o
n.
T
E
LK
O
M
N
I
K
A
I
ndones
i
an
J
our
n
al
o
f
E
l
e
c
t
r
i
c
al
E
ng
i
neer
i
ng
.
2
01
3;
11(
6)
:
3115
-
312
2.
[3
]
Ca
i
PP
,
I
ndhu
m
at
hi
C
,
Ca
i
YY
,
Z
heng
JM
,
G
ong
Y
, L
i
m
T
S
,
W
on
g
P
.
C
ol
l
i
s
i
on d
et
ec
t
i
o
n us
i
n
g ax
i
s
al
i
g
ned boun
di
n
g
box
es
.
S
i
m
ul
at
i
on
s
,
S
er
i
ou
s
G
am
es
and
T
h
ei
r
A
pp
l
i
c
at
i
ons
.
20
14
:
1
-
14
.
[4
]
G
ot
t
s
c
hal
l
k
S
,
Li
n M
C
,
M
anoc
ha D
.
O
BBT
re
e
,
A
hi
er
ar
c
hi
c
al
s
t
r
u
c
t
ur
e
f
or
r
a
pi
d
i
nt
er
f
er
e
nc
e det
e
c
t
i
on
.
P
r
oc
ee
di
n
gs
of
t
h
e 23r
d
A
n
n
ual
C
onf
er
en
c
e on C
o
m
put
er
G
r
aphi
c
s
and I
n
t
er
a
c
t
i
v
e T
ec
hni
que
s
.
1996
:
171
-
180.
[5
]
K
l
os
ow
s
k
i
JT
,
He
l
d
M
,
M
i
tc
h
e
l
l
J
SB
,
So
w
i
z
ra
l
H
, Zi
k
a
n
K
.
E
f
f
i
c
ie
n
t
c
ol
l
is
io
n
det
e
c
t
i
on us
i
ng
boundi
n
g
v
ol
um
e
hi
er
ar
c
h
i
es
of
k
-
do
ps
.
I
E
E
E
T
r
an
s
ac
t
i
on
s
o
n V
i
s
ua
l
i
z
at
i
on an
d C
om
put
er
G
r
a
phi
c
s
.
199
8
;
4
(1
):
2
1
-
37
.
[6
]
P
al
m
er
IJ
,
G
r
im
s
da
l
e
RL
.
C
o
l
l
i
s
i
on
d
et
ec
t
i
o
n
f
or
an
i
m
at
i
o
n
us
i
n
g
s
p
her
e
-
tr
e
e
s
.
C
om
put
er
G
r
aphi
c
s
F
or
um
.
199
5
;
1
4(
2)
:
105
-
11
6
.
[7
]
F
i
gue
i
r
ed
o
M
,
F
eenand
o
T
.
A
n ef
f
i
c
i
e
nt
p
ar
al
l
el
c
ol
l
i
s
i
o
n
det
e
c
t
i
o
n al
gor
i
t
hm
f
or
v
i
r
t
u
al
pr
ot
ot
y
p
e
env
i
r
onm
ent
s
.
P
r
o
c
ee
di
ng
s
of
t
he
10
t
h
I
nt
er
nat
i
on
al
C
on
f
er
e
nc
e
o
n
P
ar
a
l
l
e
l
and
D
i
s
t
r
i
but
e
d
S
y
s
t
em
s
.
2004
:
249
-
256
.
[8
]
M
ac
i
el
A
,
B
o
u
lie
R
,
T
hal
m
a
nn
D
.
E
f
f
i
c
i
ent
c
o
l
l
i
s
i
o
n
d
et
e
c
t
i
o
n
w
i
t
hi
n
def
or
m
i
ng
s
pher
i
c
al
s
l
i
di
n
g
c
ont
ac
t
.
I
EEE T
ra
n
s
a
c
t
i
o
n
s
o
n
Vi
s
u
al
i
z
at
i
on and C
om
pu
t
er
G
r
aphi
c
s
.
200
7
;
1
3(
3)
:
518
-
52
9
.
[9
]
Lai
KC
,
K
an
g
SC
.
C
o
l
l
i
s
i
on
det
e
c
t
i
o
n
s
t
r
at
eg
i
e
s
f
or
v
i
r
t
ua
l
c
o
ns
t
r
uc
t
i
on
s
i
m
ul
at
i
o
n.
A
ut
om
at
i
on i
n
C
ons
t
r
u
c
t
i
on
.
2009
;
18(
6)
:
7
24
-
736
.
[1
0
]
C
hang
JW
,
W
ang
WP
,
Ki
m
MS
.
E
f
f
i
c
i
e
nt
c
o
l
l
i
s
i
o
n det
ec
t
i
on us
i
ng a d
ual
O
B
B
-
s
pher
e
bound
i
n
g
v
ol
um
e hi
er
ar
c
hy
.
C
om
pu
t
er
A
i
ded D
e
s
i
gn
.
2
010
;
42(
1)
:
50
-
57
.
[1
1
]
C
hr
i
s
t
er
E
.
R
e
al
-
t
i
m
e c
ol
l
i
s
i
on det
e
c
t
i
o
n.
M
or
gan
K
auf
m
a
nn P
ubl
i
s
her
s
I
n
c
.
2005
.
[1
2
]
B
ade
A
,
Pi
n
g
CS
,
T
anal
ol
SH
.
C
ol
l
i
s
i
on d
et
e
c
t
i
o
n
f
or
c
l
ot
h s
i
m
ul
at
i
on u
s
i
n
g bou
ndi
ng s
ph
er
e
hi
er
ar
c
hy
.
J
ur
na
l
T
e
k
no
l
ogi
.
2
014
;
7
5
(2
):
1
-
5
.
[1
3
]
S
c
hw
es
i
ng
er
U
,
S
i
egw
ar
t
R
,
F
ur
gal
e
P
.
F
as
t
c
ol
l
i
s
i
on
det
ec
t
i
on t
hr
oug
h b
ound
i
ng
v
o
l
um
e
hi
er
ar
c
hi
e
s
i
n w
or
k
s
p
ac
e
-
t
i
m
e
s
pa
c
e
f
or
s
am
pl
i
ng
-
bas
ed m
ot
i
on
pl
a
nner
s
.
P
r
oc
e
ed
i
ng
s
of
t
he
2
015
I
E
E
E
I
nt
er
na
t
i
o
nal
C
o
nf
er
e
nc
e on R
obot
i
c
s
and
A
ut
o
m
at
i
on (
I
C
R
A
)
.
2015
:
63
-
68
.
[1
4
]
W
u
HY
,
Sh
u
Z
M
, L
i
u
YG
.
S
t
u
dy
bas
e
d o
n hy
br
i
d bo
und
i
ng
v
ol
um
e
hi
er
ar
c
hy
f
or
c
ol
l
i
s
i
on
det
e
c
t
i
o
n i
n
t
he v
i
r
t
ua
l
m
ani
pul
at
or
.
A
ppl
i
e
d M
ec
hani
c
s
a
nd M
at
er
i
al
s
.
201
3
;
4
54
:
74
-
77
.
[1
5
]
A
r
babi
E
,
B
o
u
lic
R
,
T
hal
m
an
n
D
.
F
as
t
c
o
l
l
i
s
i
o
n det
e
c
t
i
on
m
et
ho
ds
f
or
j
o
i
nt
s
ur
f
a
c
es
.
J
our
n
al
o
f
B
i
om
ec
ha
ni
c
s
.
200
9
;
42(
2)
:
9
1
-
99
.
[1
6
]
G
u
o
AB
,
W
an
g
Q
Z
,
L
i
XL
.
R
es
ear
c
h
on
c
o
l
l
i
s
i
o
n
det
e
c
t
i
on a
l
gor
i
t
hm
o
f
t
a
nk
i
n
v
i
r
t
ual
bat
t
l
ef
i
el
d
.
P
r
oc
ee
di
n
gs
o
f
I
nt
er
nat
i
on
al
C
onf
er
e
nc
e
on A
ut
o
m
at
i
on
.
20
15
:
194
4
-
19
49
.
[1
7
]
Du
P
,
Z
hao
JY
,
Pa
n
WB
,
W
ang
YG
.
G
P
U
ac
c
el
er
at
e
d
r
eal
-
t
im
e
c
o
ll
is
i
on
h
a
nd
l
in
g i
n
v
i
r
t
u
al
d
i
sa
ss
e
mb
l
y
.
J
o
ur
na
l
of
C
om
put
er
S
c
i
en
c
e
and T
ec
hnol
ogy
.
2015
;
30(
3)
:
51
1
-
518
.
[1
8
]
T
ang
M
,
M
anoc
ha
D
,
To
n
g
RF
.
M
CCD:
m
u
l
t
i
-
c
or
e c
ol
l
i
s
i
o
n
det
ec
t
i
on
bet
w
een def
or
m
a
bl
e m
od
el
s
us
i
n
g f
r
o
nt
-
ba
s
ed
de
c
om
po
s
i
t
i
on.
G
r
aph
i
c
al
M
odel
s
.
201
0
;
7
2
(2
):
7
-
23
.
[1
9
]
Xi
e
K
,
Y
ang
J
,
Z
hu
YM
.
F
as
t
c
ol
l
i
s
i
on
d
et
e
c
t
i
o
n
ba
s
ed
on
no
s
e
a
ug
m
ent
at
i
o
n
v
i
r
t
ua
l
s
ur
ger
y
.
C
om
put
er
M
et
hods
and P
r
ogr
am
s
i
n B
i
om
edi
c
i
n
e
.
2
007
;
88(
1)
:
1
-
7
.
[2
0
]
S
hen X
L,
W
u
Q
,
C
hen
g Y
W
.
H
y
br
i
d C
ol
l
i
s
i
on D
et
ec
t
i
o
n A
l
gor
i
t
h
m
ba
s
ed o
n I
m
ag
e S
pac
e.
T
E
LK
O
M
N
I
K
A
I
ndone
s
i
a
n J
ou
r
nal
o
f
E
l
e
c
t
r
i
c
al
E
ngi
neer
i
ng
.
2013;
11(
1
2)
:
7
159
-
7
165.
[2
1
]
C
hang
JW
, K
i
m
MS
. E
ffi
c
i
e
n
t tr
i
a
n
g
l
e
-
t
r
i
angl
e i
nt
er
s
e
c
t
i
on
t
es
t
f
or
O
B
B
-
bas
e
d c
ol
l
i
s
i
on
det
ec
t
i
on
.
C
om
put
er
s
and
G
r
a
phi
c
s
.
200
9
;
33(
3
)
:
23
5
-
2
40
.
[2
2
]
S
puy
R.
A
dv
anc
e
d G
am
e D
es
i
gn w
i
t
h F
l
a
s
h
.
A
pr
es
s
.
2010
:
22
4
-
236
.
[2
3
]
Ma
n
RR
,
Z
hou
DS
,
Z
han
g
Q
.
A
n
i
m
pr
ov
ed
c
ol
l
i
s
i
on
det
e
c
t
i
on
a
l
gor
i
t
hm
ba
s
ed
on
O
B
B
.
C
om
put
er
M
odel
l
i
ng a
nd N
ew
T
ec
h
nol
og
i
es
.
201
4
;
18(
1)
:
7
1
-
79
.
[2
4
]
H
ahn
JK
.
R
eal
i
s
t
i
c
ani
m
at
i
on
o
f
r
i
gi
d bo
di
e
s
.
C
om
put
er
G
r
ap
hi
c
s
.
1
988
;
22(
4)
:
299
-
308
.
[2
5
]
M
ol
l
er
T
. A
fa
s
t tr
i
a
n
g
l
e
-
t
r
i
ang
l
e i
nt
er
s
e
c
t
i
on t
e
s
t
.
J
o
ur
nal
of
G
r
aphi
c
s
T
o
ol
s
.
1
997
;
2(
2)
:
25
-
30
.
Evaluation Warning : The document was created with Spire.PDF for Python.