I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
40
,
No
.
1
,
Octo
b
er
2
0
2
5
,
p
p
.
3
2
7
~
3
4
5
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/ijeecs.v
40
.i
1
.
pp
327
-
3
4
5
327
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
The
ro
a
d conditio
ns dete
ction using
t
he
co
nv
o
lutiona
l neural
network
Su
j
it
t
ra
Sa
-
ng
iem
1
,
K
w
a
nk
a
m
o
n Dit
t
a
k
a
n
2
,
Sa
ro
ch
B
o
o
n
s
iripa
nt
3
1
C
o
l
l
e
g
e
o
f
D
i
g
i
t
a
l
S
c
i
e
n
c
e
,
P
r
i
n
c
e
o
f
S
o
n
g
k
l
a
U
n
i
v
e
r
si
t
y
,
S
o
n
g
k
h
l
a
,
T
h
a
i
l
a
n
d
2
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
i
n
g
,
P
r
i
n
c
e
o
f
S
o
n
g
k
l
a
U
n
i
v
e
r
si
t
y
,
P
h
u
k
e
t
C
a
m
p
u
s
,
P
h
u
k
e
t
,
T
h
a
i
l
a
n
d
3
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
K
a
s
e
t
s
a
r
t
U
n
i
v
e
r
si
t
y
,
B
a
n
g
k
o
k
,
T
h
a
i
l
a
n
d
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
19
,
2
0
2
4
R
ev
is
ed
J
u
n
3
,
2
0
2
5
Acc
ep
ted
J
u
l
3
,
2
0
2
5
P
o
o
r
ro
a
d
c
o
n
d
it
io
n
s
p
re
se
n
t
c
o
n
s
id
e
ra
b
le
o
b
sta
c
les
fo
r
i
n
d
iv
i
d
u
a
ls,
re
su
lt
in
g
in
a
ss
e
t
lo
ss
,
b
o
d
i
ly
h
a
rm
,
a
n
d
t
i
m
e
in
e
fficie
n
c
y
.
Ap
p
ro
x
ima
tely
1
.
3
5
m
il
li
o
n
fa
talit
ies
a
re
a
tt
rib
u
ta
b
le
to
ro
a
d
t
ra
ffic
in
c
id
e
n
ts.
T
h
e
De
p
a
rtme
n
t
o
f
P
u
b
li
c
Wo
rk
s
a
n
d
To
w
n
&
Co
u
n
tr
y
P
la
n
n
i
n
g
c
o
n
d
u
c
ted
ro
a
d
su
rv
e
y
s
to
a
ss
e
ss
a
n
d
stra
teg
ize
m
a
in
ten
a
n
c
e
e
ffo
rts.
Th
e
m
a
n
u
a
l
c
a
r
su
rv
e
y
re
q
u
ires
a
d
d
it
io
n
a
l
ti
m
e
a
n
d
a
n
e
x
c
e
ss
iv
e
b
u
d
g
e
t.
T
h
e
a
u
to
m
a
ted
sy
ste
m
o
f
a
rti
ficia
l
in
telli
g
e
n
c
e
(AI)
is
wi
d
e
ly
re
c
o
g
n
ize
d
.
Th
is
p
a
p
e
r
p
re
se
n
ts
a
m
o
d
e
l
to
d
e
tec
t
ro
a
d
su
rfa
c
e
c
o
n
d
i
ti
o
n
s
u
ti
l
izin
g
v
i
d
e
o
d
a
ta
.
F
o
u
r
v
e
rsio
n
s
of
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
)
we
re
u
ti
li
z
e
d
fo
r
th
is
wo
r
k
.
T
h
e
m
o
d
e
l
e
v
a
lu
a
ti
o
n
e
m
p
lo
y
e
d
t
h
e
m
e
a
n
a
v
e
ra
g
e
p
re
c
isio
n
(m
AP)
m
e
a
su
re
.
Th
e
v
id
e
o
d
a
ta
wa
s
a
c
q
u
ired
v
ia
a
sm
a
rtp
h
o
n
e
m
o
u
n
ted
in
a
v
e
h
icle
,
c
o
m
p
risin
g
1
0
,
9
8
4
p
h
o
to
s
fo
r
trai
n
in
g
a
n
d
2
,
1
9
8
ima
g
e
s
f
o
r
tes
ti
n
g
.
We
train
e
d
a
n
d
e
v
a
l
u
a
ted
f
o
u
r
v
e
rsio
n
s
of
CNN
a
rc
h
it
e
c
tu
re
s
n
a
m
e
d
YO
LO
,
u
ti
li
z
in
g
o
u
r
d
a
ta
a
n
d
G
P
U,
with
a
sp
e
c
ifi
c
e
m
p
h
a
sis
o
n
id
e
n
ti
fy
in
g
c
ra
c
k
s,
p
o
t
h
o
les
,
a
n
d
th
e
c
o
n
d
i
ti
o
n
o
f
m
a
n
h
o
le
c
o
v
e
rs.
Of
th
e
a
rc
h
it
e
c
tu
re
s
e
v
a
lu
a
ted
,
YO
LO
V
6
a
tt
a
in
e
d
t
h
e
g
re
a
tes
t
m
AP
sc
o
re
in
c
o
m
p
a
riso
n
YO
LO
V5
t
o
YO
LO
V8
.
T
h
e
tes
ti
n
g
re
su
lt
s
with
b
a
tch
siz
e
s
o
f
4
,
8
,
1
6
,
a
n
d
3
2
a
re
e
ffe
c
ti
v
e
.
Th
e
b
a
tch
siz
e
o
f
3
2
y
ield
s
t
h
e
h
i
g
h
e
st
p
e
rfo
rm
a
n
c
e
,
a
c
h
iev
in
g
8
7
.
3
8
%
m
AP.
C
o
n
d
u
c
t
th
e
d
ro
p
o
u
t
n
o
rm
a
li
z
a
ti
o
n
u
si
n
g
ra
te
s
o
f
0
.
2
5
,
0
.
5
0
,
0
.
7
5
,
a
n
d
1
.
T
h
e
m
a
x
imu
m
m
AP
is
o
b
se
rv
e
d
wit
h
a
d
r
o
p
o
u
t
ra
te
o
f
0
.
2
5
,
y
ield
i
n
g
a
m
AP
o
f
8
5
.
4
0
%
.
Th
e
m
o
d
e
l
i
n
d
ica
tes
th
a
t
t
h
e
g
o
v
e
rn
m
e
n
t
c
o
n
d
u
c
ted
r
o
a
d
su
rfa
c
e
i
n
sp
e
c
ti
o
n
s
with
e
n
h
a
n
c
e
d
e
fficie
n
c
y
,
e
n
a
b
li
n
g
th
e
p
lan
n
in
g
o
f
ro
a
d
re
p
a
irs
fo
r
p
u
b
li
c
u
ti
li
t
y
issu
e
s,
wh
ich
c
a
n
l
o
we
r
tr
a
n
sp
o
rtati
o
n
c
o
sts.
Ad
d
it
i
o
n
a
ll
y
,
th
e
m
o
d
e
l
c
a
n
b
e
u
ti
li
z
e
d
to
i
d
e
n
ti
f
y
h
a
z
a
rd
o
u
s
r
o
a
d
c
o
n
d
it
i
o
n
s
a
n
d
m
in
imiz
e
v
e
h
icu
lar
a
c
c
id
e
n
t
ra
tes
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
Ob
ject
d
etec
tio
n
R
o
ad
d
etec
tio
n
R
o
ad
s
u
r
f
ac
e
co
n
d
itio
n
R
o
ad
s
u
r
f
ac
e
d
etec
tio
n
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Kwa
n
k
am
o
n
Dittak
an
C
o
lleg
e
o
f
C
o
m
p
u
tin
g
,
Prin
ce
o
f
So
n
g
k
la
Un
iv
er
s
ity
,
Ph
u
k
et
C
am
p
u
s
Ph
u
k
et,
8
3
1
2
0
Kath
u
,
T
h
ailan
d
E
m
ail:
k
wan
k
am
o
n
.
d
@
p
h
u
k
et
.
p
s
u
.
ac
.
th
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
r
o
a
d
f
u
n
ctio
n
s
as
an
av
e
n
u
e
ca
p
a
b
le
o
f
p
r
o
p
ellin
g
an
ec
o
n
o
m
y
ah
ea
d
.
A
s
m
o
o
th
r
o
ad
s
u
r
f
ac
e
f
ac
ilit
ates
f
aster
an
d
m
o
r
e
ef
f
icien
t
tr
av
el.
R
o
ad
tr
a
f
f
ic
in
cid
en
ts
r
esu
lt
i
n
1
.
3
5
m
i
llio
n
f
atalities
[
1
]
.
Un
d
o
u
b
te
d
ly
,
r
o
a
d
s
ar
e
im
p
o
r
tan
t
to
s
u
s
tain
in
g
th
e
ec
o
n
o
m
y
an
d
th
e
liv
elih
o
o
d
s
o
f
in
d
iv
id
u
als.
T
h
is
en
g
en
d
e
r
s
ec
o
n
o
m
ic
an
d
c
u
ltu
r
al
ad
v
a
n
ce
m
en
t,
en
h
a
n
cin
g
n
u
m
er
o
u
s
f
ac
ets
,
in
clu
d
in
g
s
h
ip
p
in
g
,
tr
a
v
el,
co
m
m
u
n
icatio
n
,
an
d
o
th
er
s
wi
th
in
th
e
n
atio
n
.
No
n
eth
eless
,
i
f
th
e
r
o
a
d
s
u
r
f
ac
e
is
d
eter
i
o
r
at
ed
an
d
d
an
g
e
r
o
u
s
,
it
im
p
ed
es
tr
af
f
ic
a
n
d
e
n
d
an
g
e
r
s
th
e
e
n
tire
tr
an
s
p
o
r
tatio
n
s
y
s
tem
.
T
h
is
p
r
o
b
lem
lead
s
t
o
th
e
lo
s
s
o
f
b
o
th
ass
ets
an
d
life
.
Mo
r
e
o
v
er
,
th
e
q
u
ality
o
f
th
e
r
o
ad
s
u
r
f
ac
e
is
ess
en
tial
f
o
r
m
ain
tain
in
g
s
af
e
d
r
iv
in
g
,
esp
ec
ially
d
u
r
i
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
40
,
No
.
1
,
Octo
b
er
20
25
:
327
-
3
4
5
328
in
clem
en
t
wea
th
er
.
Mo
is
t
r
o
ad
s
u
r
f
ac
es
r
ed
u
ce
tire
tr
ac
tio
n
,
th
er
eb
y
in
cr
ea
s
in
g
th
e
lik
eli
h
o
o
d
o
f
ac
cid
en
ts
.
T
h
e
co
n
d
itio
n
o
f
th
e
r
o
ad
is
c
r
u
cial
in
ass
ess
in
g
th
e
s
af
ety
o
f
h
u
m
an
s
an
d
th
eir
b
elo
n
g
in
g
s
[
2
]
.
E
x
a
m
p
les
o
f
r
o
ad
c
o
n
d
itio
n
s
in
cl
u
d
e
cr
ac
k
s
,
p
o
th
o
les,
an
d
m
a
n
h
o
le
c
o
v
er
s
.
Gen
er
ally
,
th
e
d
esig
n
ated
p
eo
p
le
wo
u
ld
co
n
d
u
ct
a
m
an
u
al
i
n
s
p
ec
tio
n
o
f
th
e
r
o
ad
to
e
v
alu
ate
its
co
n
d
itio
n
.
Nev
er
th
eless
,
d
ev
el
o
p
m
en
ts
in
tech
n
o
lo
g
y
h
av
e
en
ab
le
d
ca
m
er
as
an
d
s
en
s
o
r
s
m
o
u
n
t
ed
o
n
v
eh
icles
to
g
ath
er
v
i
d
eo
f
o
o
tag
e
an
d
cr
u
cial
d
ata.
T
h
e
p
er
s
o
n
n
el
th
er
ea
f
te
r
co
n
d
u
ct
a
m
an
u
al
v
er
if
icatio
n
o
f
th
e
o
b
tain
ed
v
id
eo
m
ate
r
ial.
Desp
ite
b
ein
g
tim
e
-
co
n
s
u
m
in
g
,
th
is
p
r
o
ce
s
s
is
cr
u
cial
f
o
r
th
e
s
af
ety
o
f
liv
es a
n
d
ass
ets [
3
]
.
C
u
r
r
en
tly
,
v
ar
i
o
u
s
m
eth
o
d
o
lo
g
ies
ar
e
u
tili
ze
d
to
ev
alu
ate
r
o
ad
s
u
r
f
ac
e
co
n
d
itio
n
s
.
Vid
eo
i
m
ag
es
an
d
ca
m
er
as
h
av
e
b
ee
n
u
tili
ze
d
f
o
r
th
is
p
u
r
p
o
s
e.
Ar
tific
ial
in
tell
ig
en
ce
(
AI
)
a
n
d
m
ac
h
in
e
lear
n
in
g
(
ML
)
en
h
a
n
ce
r
o
ad
q
u
ality
d
etec
tin
g
s
y
s
tem
s
.
T
h
e
c
o
n
v
o
lu
tio
n
a
l
n
e
u
r
al
n
etwo
r
k
(
C
NN)
is
wid
ely
em
p
lo
y
ed
in
ar
tific
ial
in
tellig
en
ce
r
esear
ch
.
C
NN
i
s
ad
ep
t
at
ev
alu
atin
g
r
o
ad
c
o
n
d
itio
n
s
u
s
in
g
im
ag
es,
s
ig
n
als,
an
d
v
id
eo
d
ata
.
Ma
n
y
r
esear
ch
p
r
o
jects
em
p
lo
y
v
id
eo
o
r
s
en
s
o
r
d
ata.
YOL
O,
a
ter
m
f
o
r
"Yo
u
On
l
y
L
o
o
k
O
n
ce
,
"
is
e
x
ten
s
iv
ely
em
p
lo
y
ed
f
o
r
r
e
al
-
tim
e
o
b
ject
d
etec
tio
n
task
s
.
C
u
r
r
en
tly
,
m
an
y
v
ar
ian
ts
o
f
YOL
O
im
p
r
o
v
e
p
er
f
o
r
m
an
ce
in
te
r
m
s
o
f
m
ea
n
av
er
ag
e
p
r
ec
is
io
n
(
m
AP
).
C
ateg
o
r
ies
o
f
ac
ci
d
en
t
f
ac
to
r
s
in
r
o
a
d
ac
ci
d
en
t
d
ata
m
a
n
ag
em
en
t
r
ep
o
r
ts
i
n
clu
d
e
r
id
e
r
,
wea
th
er
,
v
eh
icle
p
er
f
o
r
m
a
n
ce
,
a
n
d
r
o
a
d
s
u
r
f
ac
e
c
o
n
d
itio
n
s
.
T
h
e
co
llis
io
n
was
ca
u
s
ed
b
y
th
e
m
an
h
o
le
co
v
er
,
a
cr
ac
k
,
a
p
o
th
o
le,
an
d
a
b
u
m
p
e
r
,
attr
ib
u
tab
le
to
th
e
r
o
a
d
co
n
d
itio
n
s
.
T
h
e
k
in
d
o
f
r
o
ad
s
u
r
f
ac
e
af
f
ec
ts
v
eh
icle
tr
ac
tio
n
.
I
f
th
e
ir
r
eg
u
lar
ter
r
ain
g
u
ar
an
t
ee
s
s
af
e
tr
a
n
s
it,
th
e
wh
ee
ls
ca
n
m
ain
tain
tr
ac
tio
n
.
T
h
ai
lan
d
f
ac
es
m
u
ltip
le
ch
allen
g
es
r
esu
ltin
g
f
r
o
m
th
e
s
u
b
s
tan
d
ar
d
q
u
ality
o
f
r
o
ad
s
u
r
f
ac
es,
r
esu
ltin
g
in
th
e
d
ev
el
o
p
m
en
t
o
f
p
o
th
o
les
an
d
cr
ac
k
s
o
v
er
tim
e
[
4
]
,
[
5
]
.
T
h
e
m
an
h
o
le
co
v
e
r
is
s
itu
ated
o
n
th
e
r
o
ad
way
.
T
h
is
g
en
er
at
es
an
elev
atio
n
an
d
d
ep
r
ess
io
n
o
n
th
e
h
ig
h
way
r
esu
ltin
g
f
r
o
m
th
e
ascen
t
a
n
d
d
es
ce
n
t
o
f
th
e
m
an
h
o
le
co
v
er
.
A
m
u
ltit
u
d
e
o
f
o
b
ject
ty
p
es
ex
is
ts
.
T
h
e
m
an
h
o
le
co
v
er
co
m
p
r
is
es
a
s
m
all
cir
cle
an
d
a
lar
g
e
r
ec
tan
g
le.
T
h
is
co
n
f
i
n
ed
th
e
m
o
d
el'
s
u
s
e
ex
clu
s
iv
ely
to
th
e
d
ata
c
o
llec
tio
n
d
o
m
ain
.
T
h
is
wo
r
k
p
r
o
p
o
s
es
th
e
u
tili
za
tio
n
o
f
a
C
N
N
o
n
v
id
e
o
d
ata
to
cr
ea
te
a
class
if
ier
ca
p
ab
le
o
f
id
en
tify
in
g
r
o
ad
s
u
r
f
ac
e
q
u
alit
y
is
s
u
es,
in
clu
d
in
g
p
o
th
o
les,
cr
ac
k
s
,
an
d
m
an
h
o
le
co
v
er
s
.
T
h
e
r
esu
lts
o
f
th
is
r
e
s
ea
r
ch
ca
n
b
e
u
tili
ze
d
in
i)
a
s
s
is
tin
g
g
o
v
e
r
n
m
e
n
t
ag
e
n
cies
in
o
p
tim
izin
g
r
o
a
d
s
u
r
v
ey
s
an
d
ass
ess
in
g
r
o
ad
q
u
ality
m
o
r
e
e
f
f
ec
tiv
ely
,
t
h
u
s
im
p
r
o
v
in
g
r
o
ad
r
ep
air
s
tr
a
teg
ies,
ii)
r
ed
u
cin
g
tr
an
s
p
o
r
tatio
n
e
x
p
en
s
es
b
y
r
e
s
o
lv
in
g
p
u
b
lic
u
tili
ty
co
n
ce
r
n
s
v
ia
th
e
p
r
o
p
o
s
ed
m
eth
o
d
,
ii
i)
in
co
r
p
o
r
atin
g
th
e
class
if
ier
in
to
an
ap
p
licatio
n
to
p
r
o
v
id
e
r
id
e
r
aler
ts
u
p
o
n
id
en
tif
y
in
g
h
az
ar
d
o
u
s
r
o
a
d
co
n
d
itio
n
s
,
an
d
iv
)
p
o
ten
tially
d
ec
r
ea
s
in
g
ca
r
ac
cid
en
t
r
ates
th
r
o
u
g
h
th
e
ap
p
licatio
n
o
f
th
is
class
if
ier
.
T
h
e
f
r
am
ewo
r
k
o
f
th
is
en
d
ea
v
o
r
b
e
g
in
s
with
d
ata
co
llectio
n
th
r
o
u
g
h
m
o
b
ile
ap
p
li
ca
tio
n
s
th
at
in
clu
d
e
v
i
d
eo
d
at
a.
T
h
e
p
r
elim
in
ar
y
s
tag
e
en
tails
f
r
am
e
ex
tr
ac
tio
n
.
T
h
e
p
ictu
r
e
lab
elin
g
is
o
p
er
at
io
n
al.
T
h
e
co
n
clu
d
in
g
p
h
ase
in
v
o
lv
es
d
ev
elo
p
in
g
th
e
o
b
ject
d
etec
tio
n
m
o
d
el
wit
h
C
NN.
T
h
e
r
esu
lts
ar
e
co
m
p
a
r
ed
with
s
ev
er
al
iter
atio
n
s
o
f
YOL
O.
T
h
e
s
u
b
s
eq
u
e
n
t
s
ec
tio
n
s
o
f
th
is
d
o
cu
m
en
t
a
r
e
s
tr
u
ctu
r
e
d
as
f
o
llo
ws
:
i
)
i
n
tr
o
d
u
ctio
n
,
ii
)
r
el
ated
wo
r
k
d
is
cu
s
s
es
r
elev
an
t
s
tu
d
ies
;
i
ii
)
m
eth
o
d
s
,
en
co
m
p
ass
in
g
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
,
d
at
a
co
llectio
n
,
p
r
e
-
p
r
o
ce
s
s
in
g
p
r
o
ce
d
u
r
es,
an
d
d
etails
o
f
th
e
C
NN
;
iv
)
r
esu
lt
an
d
d
is
cu
s
s
io
n
,
an
d
v
)
c
o
n
cl
u
s
io
n
d
is
cu
s
s
es
th
e
f
in
al
o
b
s
er
v
atio
n
s
,
an
d
a
c
k
n
o
wled
g
m
en
ts
.
2.
RE
L
AT
E
D
WO
RK
T
h
is
s
ec
tio
n
d
is
cu
s
s
es
m
u
ltip
le
r
elev
an
t
r
esear
ch
s
tu
d
ies
co
n
ce
n
tr
ated
o
n
th
e
d
etec
tio
n
o
f
ab
n
o
r
m
al
o
b
jects
o
n
th
e
r
o
ad
way
.
R
elate
d
r
esear
ch
in
clu
d
es
m
u
ltip
le
d
o
m
ain
s
,
s
u
ch
as
r
o
ad
q
u
ality
ass
ess
m
en
t,
th
e
ass
o
ciatio
n
b
etwe
en
r
o
ad
s
p
ee
d
an
d
v
ib
r
atio
n
s
,
an
d
r
o
ad
class
if
icatio
n
.
T
h
e
s
tu
d
y
in
th
is
d
o
m
ain
em
p
lo
y
s
m
an
y
h
ar
d
war
e
ty
p
es,
p
r
im
ar
ily
ca
teg
o
r
ized
in
to
two
g
r
o
u
p
s
:
i)
s
m
ar
tp
h
o
n
es
an
d
ii)
ac
ce
ler
o
m
eter
s
.
T
h
e
g
r
am
ian
an
g
u
lar
s
u
m
m
atio
n
f
ield
(
GASF)
is
u
tili
ze
d
to
co
n
v
er
t
tr
af
f
ic
tim
e
-
s
er
ies
d
ata
in
to
an
im
ag
e
f
o
r
m
at,
r
esu
ltin
g
in
d
ec
r
ea
s
in
g
th
e
er
r
o
r
r
ate
[
6
]
.
T
h
e
s
tu
d
y
ex
tr
ac
ted
d
ata
f
r
o
m
s
in
g
le
-
ax
is
an
d
th
r
ee
-
ax
is
ac
ce
ler
o
m
eter
s
.
Ma
ch
in
e
lear
n
in
g
an
d
d
ee
p
n
eu
r
al
n
etwo
r
k
s
wer
e
u
tili
ze
d
.
T
h
e
ev
alu
atio
n
o
f
class
if
icatio
n
p
er
f
o
r
m
an
ce
em
p
lo
y
s
m
an
y
p
ar
am
eter
s
ets
.
T
h
e
d
ata
co
llectio
n
u
s
in
g
an
iPh
o
n
e
6
.
T
h
r
ee
ca
teg
o
r
ies
o
f
v
eh
icles
wer
e
u
tili
ze
d
:
i)
Fo
r
d
Fo
cu
s
s
ed
an
,
ii)
Fo
r
d
Fo
cu
s
h
atch
b
ac
k
,
an
d
iii)
Su
b
ar
u
Ou
tb
ac
k
SUV.
T
h
e
Vib
r
atio
n
R
ec
o
r
d
er
ap
p
licatio
n
u
tili
ze
s
ac
ce
ler
o
m
eter
d
ata
an
d
v
id
eo
ca
p
tu
r
e
with
DJI
Osmo
.
T
h
e
r
esu
lts
in
d
icate
d
th
at
u
tili
zin
g
all
th
r
ee
ax
es
o
f
th
e
ac
ce
ler
o
m
eter
p
r
o
v
id
ed
m
o
r
e
p
r
ec
is
e
o
u
tco
m
es
co
m
p
ar
ed
to
em
p
lo
y
in
g
a
s
in
g
le
ax
is
[
7
]
.
Utilized
p
r
e
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
to
r
em
o
v
e
n
o
is
e
in
d
ata
p
r
ep
ar
atio
n
.
Su
b
s
eq
u
en
tly
,
f
ea
tu
r
e
ex
tr
ac
tio
n
was c
o
n
d
u
cted
,
f
o
llo
wed
b
y
th
e
p
r
o
m
o
tio
n
o
f
p
r
ed
ictiv
e
an
aly
s
is
.
T
h
e
r
an
d
o
m
f
o
r
est
(
R
F)
an
d
d
ec
is
io
n
tr
ee
(
DT
)
alg
o
r
ith
m
s
ar
e
em
p
lo
y
ed
f
o
r
class
if
icatio
n
an
d
th
e
id
en
tific
atio
n
o
f
p
av
em
en
t
d
is
tr
ess
k
in
d
s
[
8
]
.
A
m
ac
h
in
e
lear
n
in
g
m
o
d
el
u
tili
zin
g
a
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
was
d
ev
elo
p
ed
[
9
]
.
T
h
e
co
n
d
itio
n
o
f
th
e
r
o
ad
p
av
em
en
t
s
u
r
f
ac
e
is
ev
alu
ated
b
y
v
ib
r
atio
n
m
ea
s
u
r
em
en
ts
.
T
h
e
f
in
d
in
g
s
d
em
o
n
s
tr
ated
9
3
%
ac
cu
r
ac
y
f
o
r
th
e
R
F
m
o
d
el,
9
0
%
f
o
r
th
e
DT
m
o
d
el,
an
d
9
6
%
f
o
r
th
e
SVM
m
o
d
el.
T
h
e
U
-
ty
p
e
d
ee
p
lear
n
in
g
im
ag
e
s
e
g
m
en
tatio
n
m
o
d
el,
k
n
o
wn
as R
C
NN
-
UNe
t,
is
u
tili
ze
d
to
ex
tr
ac
t
ce
n
ter
lin
es
an
d
id
en
tify
r
o
ad
way
s
.
T
h
e
f
in
d
in
g
s
in
d
icate
a
co
m
p
leten
ess
(
C
OM
)
s
co
r
e
o
f
0
.
9
8
7
1
,
a
co
r
r
ec
tn
ess
(
C
OR
)
s
co
r
e
o
f
0
.
9
9
5
9
,
a
q
u
ality
(
Q)
s
co
r
e
o
f
0
.
9
7
4
4
,
an
d
an
F1
s
co
r
e
o
f
0
.
9
8
7
6
[
1
0
]
.
T
h
e
d
ee
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Th
e
r
o
a
d
co
n
d
itio
n
s
d
etec
tio
n
u
s
in
g
th
e
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
S
u
jitt
r
a
S
a
-
n
g
iem
)
329
r
esid
u
al
U
-
Net
is
u
tili
ze
d
to
d
elin
ea
te
th
e
r
o
ad
ar
ea
in
an
ae
r
ial
im
ag
e
[
1
1
]
.
Pre
v
io
u
s
s
tu
d
ies
o
n
r
o
ad
co
n
d
itio
n
d
etec
tio
n
h
av
e
h
ig
h
lig
h
ted
th
e
im
p
o
r
tan
ce
o
f
d
ev
ices
f
o
r
r
ec
o
r
d
in
g
an
d
s
to
r
in
g
v
id
eo
d
ata.
A
v
ar
iety
o
f
d
ev
ices h
av
e
b
ee
n
em
p
lo
y
ed
f
o
r
r
o
ad
co
n
d
itio
n
an
aly
s
is
.
T
h
e
id
en
tific
atio
n
o
f
p
o
th
o
les
h
as
b
ee
n
ac
h
iev
ed
u
s
in
g
th
er
m
al
ca
m
er
as,
as
ev
id
en
ce
d
b
y
th
e
r
esear
ch
co
n
d
u
cted
b
y
B
h
atia
et
a
l.
[
1
2
]
.
T
h
ese
ca
m
er
as,
wh
en
in
teg
r
ated
with
a
R
asp
b
er
r
y
Pi,
p
r
o
v
id
e
a
co
llectio
n
o
f
d
ata
f
o
r
th
e
id
en
tific
atio
n
o
f
p
o
th
o
les
an
d
b
u
m
p
s
.
Sm
ar
tp
h
o
n
es
h
av
e
g
ain
ed
p
o
p
u
lar
ity
as
an
ef
f
ec
tiv
e
o
p
tio
n
f
o
r
r
o
ad
d
ata
co
llectin
g
,
d
u
e
to
th
eir
h
ig
h
-
r
eso
lu
tio
n
ca
m
er
as
an
d
ac
c
ess
ib
ilit
y
.
Vid
eo
d
ata
ac
q
u
ir
ed
f
r
o
m
ce
llp
h
o
n
es
h
as
b
ee
n
u
tili
ze
d
to
id
en
tify
v
ar
io
u
s
r
o
ad
co
n
d
itio
n
s
,
s
u
ch
as
cr
ac
k
ed
s
u
r
f
ac
es,
s
m
o
o
th
r
o
ad
s
,
u
n
ev
en
ter
r
ain
,
p
o
th
o
les,
r
u
m
b
le
s
tr
ip
s
,
an
d
wate
r
[
1
2
]
.
Nu
m
er
o
u
s
s
tu
d
ies
h
av
e
u
tili
ze
d
s
m
ar
tp
h
o
n
es
to
co
llect
d
a
ta
f
o
r
th
e
d
etec
tio
n
o
f
r
o
ad
co
n
d
itio
n
s
.
T
h
ese
d
ev
ices
p
lay
a
cr
u
cial
r
o
le
in
th
e
co
llectio
n
o
f
v
id
eo
d
ata.
R
esear
ch
er
s
an
d
p
r
ac
titi
o
n
er
s
ev
alu
ate
r
o
ad
co
n
d
itio
n
s
an
d
d
ev
elo
p
m
eth
o
d
s
f
o
r
m
o
n
ito
r
in
g
an
d
id
en
tify
in
g
r
o
ad
s
u
r
f
ac
e
p
r
o
b
lem
s
.
T
h
e
C
NN
is
a
ty
p
e
o
f
b
io
-
in
s
p
ir
ed
n
eu
r
al
n
etwo
r
k
s
p
ec
if
ically
en
g
in
ee
r
ed
to
r
ep
licate
h
u
m
an
v
is
io
n
an
d
id
en
tify
o
b
jects.
C
NN
is
p
r
im
ar
ily
em
p
lo
y
ed
f
o
r
ad
d
r
ess
in
g
im
ag
e
-
r
elate
d
is
s
u
es.
T
h
e
f
u
n
d
am
en
tal
p
r
in
cip
le
o
f
C
NN
is
th
e
em
p
lo
y
m
en
t
o
f
co
n
v
o
lu
tio
n
al
lay
er
s
to
ex
tr
ac
t
ch
ar
ac
ter
is
tics
f
r
o
m
im
ag
es.
T
h
e
r
esu
ltan
t
m
o
d
el
is
q
u
alif
ied
to
m
ak
e
p
r
ec
is
e
p
r
ed
ictio
n
s
.
C
NNs
d
if
f
er
f
r
o
m
n
eu
r
al
n
etwo
r
k
s
(
NN)
b
y
th
eir
ab
ilit
y
to
ef
f
ec
tiv
ely
h
an
d
le
co
m
p
lex
m
u
ltip
le
d
atasets
,
esp
ec
ially
o
n
es
co
n
s
is
tin
g
o
f
im
ag
es.
T
h
is
ap
p
r
o
ac
h
ef
f
icien
tly
m
itig
ates
th
e
is
s
u
e
o
f
d
ata
v
ar
iatio
n
,
wh
er
ein
th
e
m
o
d
el
en
co
u
n
ter
s
d
if
f
icu
lties
in
p
r
ed
ictin
g
u
n
s
ee
n
d
ata.
C
NN
s
u
r
p
ass
es
NN
in
im
ag
e
class
if
icatio
n
task
s
.
T
h
e
ess
en
tial
co
m
p
o
n
en
ts
o
f
a
C
NN
co
n
s
is
t
o
f
th
e
co
n
v
o
lu
tio
n
al
lay
er
,
p
o
o
lin
g
lay
er
,
an
d
f
u
lly
co
n
n
ec
ted
lay
er
.
T
h
e
f
u
n
d
am
en
tal
co
n
ce
p
t
o
f
C
NN
is
th
e
co
n
v
o
lu
tio
n
al
lay
er
,
wh
ich
is
task
ed
with
f
ea
tu
r
e
ex
tr
ac
tio
n
,
in
clu
d
in
g
th
e
d
etec
tio
n
o
f
o
b
ject
ed
g
es.
C
NN
u
tili
ze
s
co
m
p
lex
m
ath
em
atica
l m
eth
o
d
s
an
d
th
e
p
r
in
cip
le
o
f
s
p
atial
co
n
v
o
lu
tio
n
f
o
r
im
ag
e
p
r
o
ce
s
s
in
g
.
Featu
r
e
ex
tr
ac
tio
n
is
ex
ec
u
ted
b
y
f
ilter
s
o
r
k
er
n
els,
ea
ch
d
esig
n
ed
to
ex
tr
ac
t
a
p
ar
ticu
lar
f
ea
tu
r
e
o
f
in
ter
est.
T
h
e
u
tili
za
tio
n
o
f
s
ev
er
al
f
ilter
s
f
u
r
th
er
en
h
an
ce
s
th
e
n
etwo
r
k
'
s
ca
p
ab
ilit
ies.
T
h
e
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
in
a
C
NN
p
r
o
d
u
ce
s
s
m
aller
m
atr
ices
as
o
u
tp
u
ts
.
T
h
er
ea
f
ter
,
th
e
p
o
o
lin
g
lay
er
ex
tr
ac
ts
s
ig
n
if
ican
t
in
f
o
r
m
atio
n
an
d
im
p
r
o
v
es
d
ata
p
r
o
ce
s
s
in
g
ef
f
icien
cy
.
T
h
er
e
ar
e
two
v
ar
ieties
o
f
p
o
o
lin
g
:
i)
m
ax
p
o
o
lin
g
,
wh
ich
id
en
tifie
s
th
e
m
ax
im
u
m
v
alu
e
with
in
ea
ch
g
r
id
,
an
d
ii)
av
er
ag
e
p
o
o
lin
g
.
T
h
e
f
u
lly
-
co
n
n
ec
ted
lay
er
co
n
s
titu
tes
th
e
co
n
clu
d
in
g
elem
en
t o
f
th
e
C
NN
ar
ch
itectu
r
e.
I
t lin
k
s
th
e
o
u
tp
u
t f
r
o
m
th
e
p
o
o
lin
g
an
d
co
n
v
o
lu
tio
n
lay
er
s
.
T
h
e
co
n
v
o
lu
tio
n
lay
er
o
f
ten
p
r
o
d
u
ce
s
a
th
r
ee
-
d
im
en
s
io
n
al
v
o
lu
m
e,
wh
er
ea
s
a
f
u
lly
-
co
n
n
ec
ted
lay
er
n
ec
ess
itates
a
o
n
e
-
d
im
en
s
io
n
al
v
ec
to
r
[
1
3
]
-
[
1
5
]
.
C
o
n
s
eq
u
en
tly
,
th
e
o
u
tp
u
t
o
f
th
e
p
o
o
lin
g
lay
er
is
tr
an
s
f
o
r
m
ed
in
to
a
v
ec
to
r
p
r
io
r
to
en
ter
in
g
th
e
f
u
lly
-
co
n
n
ec
ted
lay
er
.
Mo
r
eo
v
er
,
d
r
o
p
o
u
t
is
a
m
eth
o
d
em
p
lo
y
ed
to
r
eg
u
lar
ize
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
an
d
m
itig
ate
o
v
er
f
itti
n
g
co
n
c
er
n
s
.
I
t
r
an
d
o
m
ly
s
ets
a
p
r
o
p
o
r
tio
n
o
f
n
eu
r
o
n
o
u
tp
u
ts
d
u
r
in
g
tr
ain
in
g
.
E
n
s
em
b
les
ca
n
r
ed
u
ce
o
v
er
f
itti
n
g
b
y
av
er
ag
in
g
th
e
o
u
tp
u
ts
o
f
s
ev
er
al
m
o
d
els;
n
o
n
eth
eless
,
th
ey
ar
e
r
eso
u
r
ce
-
in
ten
s
iv
e,
tim
e
-
co
n
s
u
m
in
g
,
an
d
in
v
o
lv
e
th
e
m
an
ag
em
en
t
o
f
m
u
ltip
le
m
o
d
els [
3
]
,
[
1
5
]
.
Nu
m
er
o
u
s
s
tu
d
ies
h
av
e
em
p
lo
y
ed
d
ee
p
lear
n
in
g
m
eth
o
d
o
lo
g
ies
to
id
en
tify
ab
n
o
r
m
al
o
b
jects
in
v
id
eo
s
.
A
p
r
ev
alen
t
m
eth
o
d
in
v
o
lv
es
th
e
u
s
e
o
f
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
DC
NNs)
[
16]
with
th
e
Go
o
g
le
T
en
s
o
r
Flo
w
o
b
ject
d
etec
tio
n
(
GT
OD)
API
.
T
h
e
s
tu
d
y
ex
am
in
ed
f
iv
e
ca
teg
o
r
ies
o
f
r
o
ad
co
n
d
itio
n
s
:
i)
s
m
o
o
th
r
o
ad
,
ii)
u
n
ev
en
r
o
ad
,
iii)
p
o
th
o
le,
iv
)
in
clin
e,
an
d
v
)
b
u
m
p
.
Mo
r
eo
v
er
,
th
e
f
u
zz
y
alg
o
r
ith
m
ca
n
b
e
u
tili
ze
d
to
d
eter
m
in
e
th
e
s
p
ee
d
lim
it
o
n
th
e
h
ig
h
way
[
1
7
]
.
A
co
m
p
ar
ativ
e
s
tu
d
y
was
p
er
f
o
r
m
ed
co
m
p
ar
in
g
f
ea
tu
r
es
d
er
iv
ed
f
r
o
m
th
r
ee
ax
es
an
d
th
o
s
e
f
r
o
m
a
s
in
g
le
ax
is
.
T
h
e
r
esear
ch
em
p
lo
y
ed
SVM,
d
ec
is
io
n
tr
ee
s
,
an
d
n
eu
r
al
n
etwo
r
k
s
f
o
r
class
if
icatio
n
p
u
r
p
o
s
es.
T
h
e
im
ag
e
p
r
o
ce
s
s
in
g
p
ip
elin
e
en
co
m
p
ass
ed
lab
elin
g
,
f
ilter
in
g
,
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
.
Dee
p
n
eu
r
al
n
etwo
r
k
s
wer
e
u
tili
ze
d
to
ca
teg
o
r
ize
th
e
r
o
ad
co
n
d
itio
n
s
.
T
h
e
d
ata
co
llectio
n
u
tili
ze
d
th
r
ee
v
eh
icle
ty
p
es:
i)
Fo
r
d
Fo
cu
s
Sed
an
,
ii)
Fo
r
d
Fo
cu
s
Hatc
h
b
ac
k
,
an
d
iii)
Su
b
ar
u
Ou
tb
ac
k
SUV.
T
h
e
v
id
eo
r
ec
o
r
d
in
g
s
wer
e
o
b
tain
ed
with
an
iPh
o
n
e
6
.
T
h
e
p
r
im
ar
y
f
o
cu
s
o
f
th
e
r
o
ad
s
u
r
f
ac
e
s
tu
d
y
was th
e
id
en
tific
atio
n
o
f
p
o
th
o
les [
7
]
.
W
ir
atm
o
k
o
et
al.
co
n
d
u
cted
a
s
tu
d
y
to
id
en
tify
p
o
th
o
les
o
n
th
e
r
o
ad
way
.
T
h
e
s
p
ec
if
ic
cr
iter
ia
ar
e
a
d
iam
eter
ex
ce
ed
in
g
10
ce
n
tim
eter
s
an
d
a
d
ep
th
o
f
at
least
5
ce
n
tim
eter
s
.
T
h
e
wr
ap
p
in
g
an
d
cr
o
p
p
in
g
tech
n
iq
u
es
ar
e
u
tili
ze
d
f
o
r
o
b
ject
d
etec
tio
n
,
an
d
a
C
NN
b
ased
o
n
L
eNe
t
5
m
ay
p
r
o
d
u
ce
th
e
m
o
d
el.
T
h
e
f
in
d
in
g
s
in
d
icate
d
an
ac
cu
r
ac
y
o
f
9
2
.
8
%
[
1
2
]
.
I
n
a
s
ep
ar
ate
s
tu
d
y
,
a
p
o
th
o
le
d
etec
tin
g
s
y
s
tem
u
tili
zin
g
a
m
ix
tu
r
e
o
f
Gau
s
s
ian
s
(
Mo
G)
co
m
b
in
ed
w
ith
an
SVM
m
o
d
el
an
d
f
aster
R
-
C
NN.
I
n
th
e
Mo
G
m
o
d
el,
l
in
ea
r
SVM
an
d
r
ad
ial
b
asis
f
u
n
ctio
n
SVM
(
R
B
F
-
SVM)
wer
e
u
tili
ze
d
.
No
n
eth
eless
,
th
e
s
tu
d
y
o
f
th
e
v
id
eo
d
ata
in
d
icate
d
th
at
th
e
Mo
G
+
SVM
was
in
ap
p
r
o
p
r
iate.
C
o
n
v
er
s
ely
,
f
aster
R
C
NN
d
em
o
n
s
tr
ated
s
u
p
er
io
r
pe
r
f
o
r
m
an
ce
in
v
eh
icle
r
ec
o
g
n
itio
n
at
n
ig
h
t.
A
d
ee
p
b
elief
n
etwo
r
k
(
DB
N)
is
u
tili
ze
d
to
d
ev
elo
p
th
e
f
u
n
ctio
n
f
o
r
ass
ess
in
g
g
lo
b
al
r
o
ad
s
af
ety
p
er
f
o
r
m
an
ce
(
SF
P).
T
h
is
m
o
d
el
ca
n
f
o
r
ec
ast
th
e
f
r
eq
u
en
cy
o
f
ac
cid
en
ts
ac
r
o
s
s
d
iv
er
s
e
lo
ca
tio
n
s
.
T
h
e
o
u
tco
m
e
o
f
en
h
an
ce
d
p
er
f
o
r
m
an
ce
is
a
r
ed
u
ctio
n
o
f
th
e
m
ax
im
u
m
er
r
o
r
r
ate
to
2
3
.
1
5
%
[
1
8
]
.
T
h
e
r
ef
in
ed
m
ask
R
-
C
NN
(
R
M
R
-
C
NN)
u
tili
ze
s
an
en
d
-
to
-
en
d
lear
n
in
g
ap
p
r
o
ac
h
f
o
r
th
e
ac
q
u
is
itio
n
o
f
r
ea
l
-
tim
e
r
o
ad
p
ictu
r
es.
T
o
id
en
tify
th
e
tr
af
f
ic
s
ig
n
s
f
o
r
th
e
co
m
p
ar
is
o
n
o
f
f
ast
R
-
C
NN
an
d
m
ask
R
-
C
NN
[
19
]
.
T
h
e
r
an
d
o
m
f
o
r
est
r
eg
r
ess
o
r
(
R
FR
)
,
m
u
ltio
u
tp
u
t
r
eg
r
ess
o
r
(
MO
R
)
,
an
d
ar
tific
ial
n
eu
r
al
n
et
wo
r
k
(
ANN)
ca
n
b
e
d
esig
n
ed
to
ass
ess
m
o
v
em
en
t a
t
in
ter
s
ec
tio
n
s
.
T
h
e
an
aly
s
is
o
f
th
e
co
r
r
elatio
n
b
etwe
en
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
40
,
No
.
1
,
Octo
b
er
20
25
:
327
-
3
4
5
330
v
o
lu
m
es
an
d
co
m
p
atib
le
tu
r
n
in
g
m
o
v
em
en
ts
[
2
0
]
.
T
h
e
ass
ess
m
en
t
o
f
p
av
em
en
t
p
er
f
o
r
m
an
ce
in
co
r
p
o
r
ates
v
ar
io
u
s
cr
iter
ia,
in
clu
d
in
g
b
ase
ty
p
e,
h
o
t
m
ix
asp
h
alt
co
n
cr
ete
(
HM
AC
)
lay
er
th
ick
n
ess
,
b
ase
lay
er
th
ick
n
ess
,
tr
af
f
ic
lo
ad
,
ag
e,
an
d
clim
atic
co
n
d
itio
n
s
s
u
ch
as
p
r
ec
ip
itatio
n
an
d
tem
p
er
atu
r
e.
T
h
is
s
tu
d
y
em
p
lo
y
s
v
ar
io
u
s
s
tatis
tical
m
eth
o
d
o
lo
g
ies,
in
clu
d
in
g
g
en
er
alize
d
lin
ea
r
m
o
d
els
(
GL
M)
,
b
in
ar
y
lo
g
is
tic
r
eg
r
ess
io
n
(
B
L
R
)
,
an
d
r
an
d
o
m
f
o
r
est
(
R
F).
T
h
e
lo
n
g
-
ter
m
p
av
em
en
t
p
er
f
o
r
m
an
ce
p
r
o
g
r
am
(
L
T
PP
)
ex
ce
ls
in
d
ata
an
aly
s
is
.
Fu
r
th
er
m
o
r
e,
th
e
p
er
f
o
r
m
an
ce
o
f
r
eg
r
ess
io
n
m
o
d
els
(
GL
M
an
d
B
L
R
)
is
ass
ess
ed
in
r
elatio
n
to
R
F
b
y
th
e
cr
o
s
s
-
v
alid
atio
n
m
eth
o
d
.
T
h
e
f
in
d
in
g
s
r
ev
ea
l
th
e
im
p
o
r
tan
ce
o
f
d
esig
n
ch
ar
ac
ter
is
tics
,
s
u
ch
as
cu
ttin
g
,
r
o
u
g
h
n
ess
,
an
d
allig
ato
r
cr
ac
k
in
g
[
2
1
]
.
Mu
k
esh
et
a
l.
[
2
1
]
id
en
tifie
d
cr
a
ck
s
an
d
p
o
th
o
les
in
th
e
r
o
ad
b
y
co
m
p
ar
in
g
R
an
d
o
m
Fo
r
est
an
d
C
NN
m
eth
o
d
o
lo
g
ies.
T
h
e
d
em
o
n
s
tr
ated
ac
cu
r
ac
y
is
9
6
%
[
2
2
]
.
R
o
ad
d
etec
tio
n
m
ay
id
en
tify
f
ea
tu
r
es
f
r
o
m
s
ev
er
al
im
ag
e
s
o
u
r
ce
s
,
in
clu
d
in
g
s
atellite
im
ag
er
y
.
T
h
e
m
o
d
el
ca
n
b
e
co
n
s
tr
u
cted
u
s
in
g
ANN,
C
NN
alo
n
g
with
m
o
r
p
h
o
lo
g
ical
en
h
an
ce
m
en
t
an
d
s
eg
m
en
tatio
n
tech
n
iq
u
es
[
2
3
]
-
[
2
5
]
.
R
esear
ch
h
as
b
ee
n
co
n
d
u
cted
to
m
itig
ate
s
h
ad
o
ws
th
r
o
u
g
h
a
s
h
ad
o
w
-
in
v
ar
ian
t
f
ea
tu
r
e
s
p
ac
e
[
2
6
]
.
An
o
th
er
f
o
r
m
o
f
im
ag
e
u
tili
ze
d
f
o
r
r
o
ad
d
etec
tio
n
is
v
er
y
h
ig
h
r
eso
lu
tio
n
(
VHR)
r
em
o
t
e
s
en
s
in
g
.
T
h
is
im
ag
e
f
o
r
m
at
is
s
u
itab
le
f
o
r
s
eg
m
en
tatio
n
[
2
7
]
.
T
h
e
ae
r
ial
im
ag
e
ca
n
b
e
u
tili
ze
d
to
p
r
o
d
u
ce
a
m
ap
b
ased
o
n
s
em
an
tic
s
eg
m
en
tatio
n
.
An
u
n
m
an
n
ed
ae
r
ial
v
eh
icle
(
UAV)
is
em
p
lo
y
ed
to
au
to
n
o
m
o
u
s
ly
cr
ea
te
th
e
m
o
n
ito
r
in
g
r
o
ad
ar
ea
b
ased
o
n
r
ap
id
h
o
m
o
g
r
ap
h
y
,
u
tili
zin
g
a
2
D
B
ir
d
'
s
ey
e
v
iew
(
B
E
V)
im
ag
e
to
r
ec
o
g
n
ize
r
o
ad
s
,
lan
es,
an
d
d
r
iv
er
b
eh
av
io
r
,
y
ield
in
g
f
av
o
r
ab
le
r
esu
lts
[
2
8
]
-
[
3
0
]
.
T
h
e
r
ec
u
r
r
en
t
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
U
-
N
et
(
R
C
NN
-
UNe
t)
is
u
tili
ze
d
f
o
r
r
o
ad
d
etec
tio
n
an
d
ce
n
ter
lin
e
ex
tr
ac
tio
n
[
3
1
]
.
T
h
e
m
u
lti
-
ce
n
ter
ed
h
o
u
g
h
f
o
r
est
(
MCHF)
m
eth
o
d
ca
n
b
e
em
p
lo
y
ed
f
o
r
cu
r
v
ilin
ea
r
d
etec
tio
n
in
im
ag
es
as
a
r
ep
lace
m
en
t
to
e
u
clid
ea
n
d
is
tan
ce
[
3
2
]
.
R
eg
ar
d
in
g
co
lo
r
r
o
ad
d
etec
tio
n
,
en
h
an
ce
m
en
t
ca
n
b
e
ac
h
iev
ed
b
y
u
tili
zin
g
illu
m
in
an
t
in
tr
in
s
ic
im
ag
es
to
p
r
o
d
u
ce
a
p
ix
el
-
lev
el
co
n
f
id
en
ce
m
ap
[
3
3
]
.
T
h
e
d
etec
tio
n
o
f
lin
e
s
eg
m
en
tatio
n
is
ac
co
m
p
lis
h
ed
b
y
th
e
p
r
o
b
ab
ilis
tic
Ho
u
g
h
tr
an
s
f
o
r
m
,
wh
ich
d
elin
ea
tes
th
e
r
o
ad
s
ec
tio
n
in
ae
r
ial
im
ag
es
u
tili
zin
g
a
h
is
to
g
r
am
[
3
4
]
,
[
3
5
]
.
T
h
e
r
o
ad
b
o
u
n
d
ar
y
ca
n
b
e
ex
tr
ac
ted
u
s
in
g
th
e
s
-
FC
N
-
lo
c
ap
p
r
o
ac
h
,
wh
ich
o
p
er
ates
30%
f
aster
th
an
FC
N
[
3
6
]
.
An
au
to
n
o
m
o
u
s
d
etec
tio
n
s
y
s
tem
was
im
p
lem
en
ted
to
id
en
tify
d
iv
er
s
e
r
o
ad
co
n
d
itio
n
s
,
s
u
ch
as
p
o
th
o
les,
s
tr
ee
t
b
u
m
p
s
,
an
d
d
r
iv
er
b
eh
av
io
r
s
,
u
tili
zin
g
a
co
m
b
in
atio
n
o
f
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
n
eu
r
al
n
etwo
r
k
s
an
d
a
r
eser
v
o
ir
co
m
p
u
tin
g
(
R
C
)
m
o
d
el.
T
h
is
ap
p
r
o
ac
h
ac
h
iev
es
an
ac
cu
r
ac
y
o
f
0
.
9
8
in
d
is
tin
g
u
is
h
in
g
b
etwe
en
p
o
th
o
les an
d
th
eir
ab
s
en
ce
[
3
7
]
.
Pre
v
io
u
s
r
esear
ch
h
as
in
tr
o
d
u
ce
d
m
u
ltip
le
m
eth
o
d
o
lo
g
ies
f
o
r
ev
alu
atin
g
r
o
ad
s
u
r
f
ac
e
co
n
d
itio
n
s
.
Pre
-
p
r
o
ce
s
s
in
g
p
r
o
ce
s
s
es
en
co
m
p
ass
ac
tiv
ities
s
u
ch
as
lab
elin
g
,
f
ilter
in
g
,
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
.
Dee
p
lear
n
in
g
m
o
d
els s
u
ch
as YO
L
O
V
2
,
YOL
O
V
3
,
an
d
u
p
to
YOL
O
V
8
ar
e
f
r
eq
u
en
tly
u
tili
ze
d
f
o
r
th
is
o
b
jectiv
e.
I
m
ag
es c
an
b
e
co
llected
u
tili
zin
g
ce
llp
h
o
n
es a
n
d
v
id
eo
ca
m
er
as.
3.
M
E
T
H
O
D
S
T
h
is
s
ec
tio
n
ex
p
lain
s
th
e
p
r
o
c
ess
o
f
th
is
p
ap
er
.
T
h
er
e
a
r
e
f
o
u
r
s
u
b
s
ec
tio
n
s
.
Su
b
s
ec
tio
n
3
.
1
d
is
cu
s
s
es
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
o
u
t
lin
es
th
e
u
tili
ze
d
f
r
am
ewo
r
k
.
Su
b
s
ec
tio
n
3
.
2
Data
c
o
lle
ctio
n
s
p
ec
if
ies
th
e
m
eth
o
d
o
l
o
g
y
f
o
r
d
ata
ac
q
u
is
i
tio
n
.
S
ubs
ec
tio
n
3
.
3
.
Data
p
r
ep
r
o
ce
s
s
in
g
o
u
tlin
es
th
e
p
r
eli
m
in
ar
y
p
r
o
ce
d
u
r
es
b
ef
o
r
e
m
o
d
el
d
e
v
elo
p
m
e
n
t.
An
d
S
ubs
ec
tio
n
3
.
4
.
C
NN
an
aly
ze
s
th
e
h
is
to
r
ical
ev
o
lu
tio
n
o
f
C
NN
an
d
ea
ch
v
er
s
io
n
o
f
YOL
O
u
p
to
YOL
O
V8
3
.1
.
P
r
o
po
s
ed
f
ra
m
ewo
r
k
T
h
is
s
ec
tio
n
p
r
e
s
en
ts
a
f
r
am
ewo
r
k
d
esig
n
e
d
f
o
r
th
e
i
d
en
tific
atio
n
o
f
r
o
ad
s
u
r
f
ac
e
co
n
d
itio
n
s
.
Fig
u
r
e
1
d
em
o
n
s
tr
ates
th
e
f
r
a
m
ewo
r
k
.
T
h
is
en
a
b
les
th
r
ee
s
tag
es
f
o
r
ev
alu
atin
g
r
o
a
d
s
u
r
f
a
ce
co
n
d
itio
n
s
with
v
id
eo
d
ata.
T
h
e
p
r
o
ce
d
u
r
e
in
clu
d
es
d
ata
c
o
llectio
n
,
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
a
n
d
m
o
d
e
l
g
en
er
atio
n
.
T
h
e
f
r
am
ewo
r
k
i
n
itiates
b
y
g
ath
er
i
n
g
d
ata
an
d
em
p
lo
y
in
g
a
s
m
ar
tp
h
o
n
e'
s
ca
m
er
a
to
r
ec
o
r
d
v
id
eo
d
ata.
T
h
e
f
r
am
e
ex
tr
ac
tio
n
in
v
o
lv
es
co
n
v
er
tin
g
th
e
m
o
v
ie
in
to
s
ev
er
al
p
ictu
r
es.
Up
o
n
co
m
p
letio
n
o
f
th
e
d
ata
co
llectio
n
an
d
f
r
am
e
ex
tr
ac
ti
o
n
p
h
ase,
t
h
e
n
ex
t
s
tag
e
is
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
.
Prio
r
to
co
m
m
en
ci
n
g
th
is
p
h
ase,
d
ata
au
g
m
en
tatio
n
is
r
eq
u
i
r
ed
.
T
h
i
s
aim
s
to
im
p
r
o
v
e
th
e
d
ata
q
u
ality
.
Div
er
s
e
im
ag
e
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
ar
e
u
tili
ze
d
to
g
u
ar
an
tee
th
e
d
ata
is
f
o
r
m
atted
co
r
r
ec
tly
f
o
r
f
u
r
th
er
p
r
o
ce
s
s
in
g
.
T
h
ese
m
ec
h
an
is
m
s
en
co
m
p
ass
lan
e
d
etec
tio
n
an
d
o
b
ject
d
etec
tio
n
.
T
h
is
s
tu
d
y
co
n
ce
n
tr
ates
o
n
th
e
v
eh
icu
lar
lan
e
a
n
d
s
u
b
s
eq
u
en
tly
th
e
lan
e
d
etec
tin
g
p
r
o
ce
d
u
r
e.
Su
b
s
eq
u
e
n
t
to
th
e
en
h
an
ce
m
en
t
o
f
p
ictu
r
e
lab
elin
g
,
th
is
p
h
ase
d
elin
ea
tes
th
e
o
b
ject
f
r
om
th
e
b
ac
k
d
r
o
p
,
a
cr
u
cial
s
tep
f
o
r
en
ab
lin
g
th
e
m
ac
h
in
e
to
r
e
co
g
n
ize
an
d
d
if
f
er
en
tiate
th
e
o
b
ject.
T
h
e
o
u
tco
m
e
o
f
th
e
lab
elin
g
o
p
e
r
atio
n
is
an
XM
L
f
ile
em
p
lo
y
ed
to
co
n
s
tr
u
ct
th
e
m
o
d
el
u
tili
zin
g
th
e
C
NN.
3
.
2.
Da
t
a
c
o
llect
io
n
T
h
e
d
ata
o
n
r
o
ad
co
n
d
itio
n
,
p
ar
ticu
lar
ly
ab
o
u
t
asp
h
alt
r
o
a
d
s
,
is
u
tili
ze
d
to
d
ev
elo
p
th
e
class
if
ier
.
Data
co
llectio
n
co
n
d
u
cted
b
y
s
m
ar
tp
h
o
n
es
f
r
o
m
Ma
y
2
0
2
0
to
Sep
tem
b
er
2
0
2
1
in
Ph
u
k
et
an
d
B
an
g
k
o
k
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Th
e
r
o
a
d
co
n
d
itio
n
s
d
etec
tio
n
u
s
in
g
th
e
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
S
u
jitt
r
a
S
a
-
n
g
iem
)
331
T
h
ailan
d
.
A
b
r
ac
k
et
was
u
tili
ze
d
to
h
o
ld
t
h
e
s
m
ar
tp
h
o
n
e
tig
h
tly
in
th
e
v
e
h
icle,
as
s
ee
n
in
Fig
u
r
e
2
.
An
ap
p
licatio
n
was e
m
p
lo
y
e
d
to
s
av
e
an
d
d
ir
ec
tly
u
p
lo
ad
th
e
co
llected
d
ata
to
th
e
clo
u
d
.
Fig
u
r
e
1
.
T
h
e
r
o
a
d
s
u
r
f
ac
e
co
n
d
itio
n
s
d
etec
tio
n
f
r
a
m
ewo
r
k
Fig
u
r
e
2
.
T
h
e
s
etu
p
o
f
eq
u
ip
m
en
t
Su
b
s
eq
u
en
tly
,
f
r
am
e
ex
tr
ac
tio
n
is
ex
ec
u
ted
to
o
b
tain
im
ag
es
f
r
o
m
th
e
v
id
eo
,
with
ea
c
h
im
ag
e
s
ca
led
to
1
9
2
0
×
1
0
8
0
p
ix
els.
T
h
e
d
ataset
co
m
p
r
is
es
1
0
,
9
8
4
im
ag
es
d
esig
n
ated
f
o
r
tr
ain
in
g
an
d
2
,
1
9
8
im
ag
es
allo
ca
ted
f
o
r
test
in
g
.
T
ab
le
1
p
r
esen
ts
co
m
p
r
eh
e
n
s
iv
e
in
f
o
r
m
atio
n
co
n
ce
r
n
in
g
th
e
v
i
d
eo
s
an
d
th
e
p
r
o
ce
d
u
r
e
o
f
f
r
am
e
ex
tr
ac
tio
n
.
Fig
u
r
e
3
p
r
o
v
id
es a
n
illu
s
tr
atio
n
o
f
a
cu
s
to
m
d
ataset.
T
ab
le
1
i
llu
s
tr
ates
th
e
f
r
am
e
ex
tr
ac
tio
n
s
tatis
tics
.
A
to
tal
o
f
1
3
4
,
1
7
6
s
ec
o
n
d
s
r
esu
l
ted
in
th
e
ex
tr
ac
tio
n
o
f
1
,
3
4
1
,
7
6
0
f
r
am
e
s
.
Am
o
n
g
t
h
ese
f
r
am
es,
e
v
er
y
th
r
ee
f
r
am
es
will
s
elec
t
o
n
e
f
r
am
e.
T
h
e
n
,
2
4
,
2
7
6
co
n
tain
ed
o
b
jects
wer
e
ch
o
s
e
n
f
o
r
g
en
er
atin
g
th
e
o
b
ject
d
e
tect
io
n
m
o
d
el.
An
d
th
e
f
r
am
e
ex
tr
ac
tio
n
co
d
e
is
s
h
o
wn
in
Alg
o
r
ith
m
1
.
T
h
e
v
id
eo
f
ile
is
d
esig
n
ated
as
'
v
id
eo
ca
p
'
,
th
e
im
ag
e
c
o
u
n
t
is
r
ef
er
r
ed
to
as
'
co
u
n
t'
,
an
d
th
e
s
tate
o
f
th
e
v
id
e
o
r
ea
d
is
in
d
icate
d
b
y
th
e
p
ar
am
eter
'
s
u
cc
ess
'
(
lin
es 2
-
4
)
.
Su
b
s
eq
u
e
n
tly
,
th
e
r
ec
o
r
d
e
d
v
id
e
o
is
an
aly
ze
d
,
an
d
th
e
co
u
n
t
p
a
r
am
eter
is
ex
ec
u
te
d
(
lin
es
7
-
8
)
.
T
h
e
im
ag
e
is
th
er
ea
f
ter
s
av
ed
to
th
e
d
ir
ec
to
r
y
af
ter
ev
er
y
th
r
ee
im
a
g
es (
lin
es 9
-
1
2
)
,
co
n
s
titu
tin
g
th
e
o
u
tp
u
t.
T
ab
le
1
.
T
h
e
in
tr
icac
ies o
f
v
id
eo
an
d
f
r
am
e
e
x
tr
ac
tio
n
V
i
d
e
o
Ti
me(s
)
A
l
l
F
r
a
m
e
F
r
a
me
w
i
t
h
o
b
j
e
c
t
S
e
l
e
c
t
e
d
f
r
a
me
1
2
5
1
3
4
,
1
7
6
1
,
3
4
1
,
7
6
0
3
2
,
8
0
3
2
4
,
2
7
6
3
.
3
.
Da
t
a
p
re
pro
ce
s
s
ing
Fo
llo
win
g
d
ata
co
llectio
n
u
s
in
g
a
s
m
ar
tp
h
o
n
e
ap
p
licatio
n
,
f
r
am
e
e
x
tr
ac
tio
n
is
p
er
f
o
r
m
ed
.
T
h
is
s
ec
tio
n
ad
d
r
ess
es
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
,
co
m
p
r
is
in
g
t
h
r
ee
s
u
b
s
ec
tio
n
s
.
Su
b
s
u
b
s
ec
tio
n
3
.
3
.
1
ad
d
r
ess
es
th
e
ca
teg
o
r
y
o
f
d
ata.
Nex
t
3
.
3
.
2
co
n
s
titu
tes
a
co
m
p
o
n
en
t
o
f
th
e
lan
e
d
etec
tin
g
p
r
o
ce
d
u
r
e.
S
u
b
s
u
b
s
ec
tio
n
3
.
3
.
3
d
elin
ea
tes th
e
s
p
ec
if
ics o
f
o
b
je
ct
lab
elin
g
,
a
cr
u
cial
p
r
o
ce
d
u
r
e
p
r
io
r
t
o
d
ata
in
p
u
t f
o
r
m
o
d
el
g
en
er
atio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
40
,
No
.
1
,
Octo
b
er
20
25
:
327
-
3
4
5
332
Fig
u
r
e
3
.
A
d
em
o
n
s
tr
atio
n
o
f
a
cu
s
to
m
d
ataset
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Th
e
r
o
a
d
co
n
d
itio
n
s
d
etec
tio
n
u
s
in
g
th
e
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
S
u
jitt
r
a
S
a
-
n
g
iem
)
333
Alg
o
r
ith
m
1
.
Fra
m
e
ex
tr
ac
tio
n
1: Input:
2: videocap = The video file
3: count = The image count
4: success =
The status of the video read
5:
Output:
6:
while
success do
7: success = videocap.read()
8: count = count + 1
9: If count mod 3 = 0
10: If success = true
11: cv2.imwrite("image path" % count, image, [cv2.IMWRITE_JPEG_QUALITY
, 100])
12: end if
13: end if
14:
end while
3
.
3
.1
.
Da
t
a
c
a
t
eg
o
ries
T
h
is
s
tu
d
y
id
en
tifie
s
th
r
ee
s
p
ec
if
ic
o
b
jects
f
o
r
i
n
s
p
ec
tio
n
b
y
th
e
De
p
ar
tm
en
t
o
f
Hig
h
way
s
i
n
T
h
ailan
d
.
T
h
ese
o
b
jects
in
clu
d
e
r
o
a
d
d
am
ag
e
,
class
if
ied
as
s
tr
u
ctu
r
al
f
ailu
r
e
an
d
f
u
n
c
tio
n
al
f
ailu
r
e
.
T
h
e
s
tr
u
ctu
r
e
f
ailu
r
e
ca
teg
o
r
y
en
co
m
p
ass
es
p
o
th
o
les
an
d
cr
ac
k
s
,
wh
er
ea
s
m
an
h
o
les
ar
e
class
if
ied
u
n
d
er
th
e
f
u
n
ctio
n
al
f
ailu
r
e
ca
teg
o
r
y
.
T
h
e
im
p
ac
ts
o
f
f
u
n
ctio
n
al
f
ailu
r
e
r
elate
to
co
n
ce
r
n
s
o
f
c
o
n
v
en
ien
ce
an
d
s
af
ety
.
Fig
u
r
e
4
illu
s
tr
ates
an
ex
am
p
l
e
o
f
th
ese
th
in
g
s
.
Fig
u
r
e
4
(
a)
d
ep
icts
a
cr
ac
k
,
Fig
u
r
e
4
(
b
)
il
lu
s
tr
ates
a
p
o
th
o
le,
an
d
Fig
u
r
e
4
(
c)
s
h
o
wca
s
es a
m
an
h
o
le
c
o
v
er
.
A
n
ex
am
p
le
o
f
a
cu
s
to
m
o
b
ject
is
illu
s
tr
ated
in
Fig
u
r
e
5
.
(
a)
(
b
)
(
c)
Fig
u
r
e
4
.
An
illu
s
tr
atio
n
o
f
t
h
e
th
r
ee
o
b
jects
(
a)
an
ex
a
m
p
le
o
f
a
cr
ac
k
(
b
)
a
n
ex
am
p
le
o
f
a
p
o
th
o
le
,
an
d
(
c)
an
e
x
am
p
le
o
f
a
m
an
h
o
le
3
.
3
.2
.
L
a
ne
d
et
ec
t
io
n
Fo
llo
win
g
th
e
co
llectio
n
o
f
v
id
eo
d
ata
,
in
cl
u
d
in
g
th
r
ee
o
b
j
ec
t
ty
p
es,
th
e
f
r
a
m
e
ex
tr
ac
tio
n
p
r
o
ce
s
s
co
m
m
en
ce
s
.
T
h
e
f
r
am
e
e
x
tr
a
ctio
n
tech
n
iq
u
e
in
v
o
l
v
es
co
n
v
er
tin
g
a
v
id
eo
co
llectio
n
in
to
s
ev
er
al
im
ag
es.
Imp
r
o
v
em
en
ts
wer
e
im
p
lem
e
n
ted
in
th
e
lan
e
d
etec
tio
n
p
r
o
ce
d
u
r
e,
p
ar
ticu
la
r
ly
f
o
cu
s
in
g
o
n
id
en
tify
in
g
lan
es
f
o
r
v
eh
icu
lar
d
r
iv
in
g
.
T
h
e
lan
e
d
etec
tin
g
p
h
ase
is
ess
en
tial,
in
clu
d
in
g
a
s
eq
u
en
ce
o
f
ac
tio
n
s
.
I
n
itially
,
C
an
n
y
E
d
g
e
d
etec
tio
n
is
im
p
lem
en
te
d
,
s
u
cc
ee
d
ed
b
y
ar
ea
s
eg
m
en
t
atio
n
,
an
d
u
ltima
tely
,
th
e
a
p
p
li
ca
tio
n
o
f
th
e
Ho
u
g
h
T
r
an
s
f
o
r
m
.
Fig
u
r
e
6
illu
s
tr
ates th
e
r
esu
lts
,
wh
ile
Alg
o
r
ith
m
2
d
is
p
lay
s
th
e
lan
e
d
etec
tin
g
co
d
e.
Alg
o
r
ith
m
2
elev
ates th
e
im
ag
e
co
llectio
n
to
th
e
p
ar
am
eter
'
f
r
am
e'
(
lin
e
2
)
.
Su
b
s
eq
u
en
tly
,
im
p
lem
en
t
a
C
an
n
y
m
eth
o
d
to
i
d
en
tify
ed
g
es
with
in
an
im
ag
e.
T
h
e
s
eg
m
en
tatio
n
is
ex
ec
u
ted
to
p
r
o
d
u
ce
th
e
p
ix
el
r
eg
io
n
in
an
im
ag
e.
Utilize
th
e
Ho
u
g
h
tr
an
s
f
o
r
m
t
o
id
en
tif
y
th
e
s
h
a
p
e
(
lin
es
4
-
6
)
.
T
h
e
f
in
al
s
tep
in
v
o
lv
es
ca
lcu
latin
g
th
e
lin
e
an
d
s
ee
in
g
th
e
o
u
t
p
u
t
(
lin
es 7
-
1
0
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
40
,
No
.
1
,
Octo
b
er
20
25
:
327
-
3
4
5
334
Alg
o
r
ith
m
2
.
L
an
e
d
etec
tio
n
1: Input:
2: frame = The collection of read image files
3: Output:
4: canny = do_canny(frame)
5: segment = do_segment(canny)
6: hough = HoughLinesP(segment, 2, minLineLoth = 100, maxLineGap = 50)
7: li
nes = calculate_lines(frame, hough)
8: line_visualize = visualize_lines(frame, lines)
9: output = addWeighted(frame, 0.9, line_visualize, 1, 1)
10:print output
Su
b
s
eq
u
en
t
to
th
e
ac
q
u
is
itio
n
o
f
v
id
eo
d
ata,
th
e
v
id
eo
s
wer
e
an
aly
ze
d
to
ex
tr
ac
t
d
is
cr
ete
p
ictu
r
es.
T
h
e
in
itial
d
im
en
s
io
n
s
o
f
th
ese
p
h
o
to
s
wer
e
1
9
2
0
×
1
0
8
0
p
ix
els.
Fo
r
th
e
p
u
r
p
o
s
es
o
f
o
u
r
in
v
esti
g
atio
n
,
we
r
esized
th
e
p
h
o
to
s
to
1
2
8
0
×
7
2
0
p
ix
els.
T
o
elim
in
ate
u
n
wan
ted
en
v
ir
o
n
m
en
tal
co
m
p
o
n
en
ts
,
we
ex
cised
3
6
0
p
ix
els
f
r
o
m
th
e
to
p
an
d
3
2
0
p
ix
els
f
r
o
m
b
o
th
th
e
lef
t
an
d
r
ig
h
t
s
id
es
o
f
th
e
p
h
o
to
s
.
Fig
u
r
e
7
d
ep
icts
th
e
o
r
ig
in
al
im
ag
e
alo
n
g
with
th
e
s
p
ec
if
ied
cr
o
p
r
eg
io
n
.
T
h
e
d
esig
n
ated
ar
ea
o
f
th
e
im
ag
e,
m
ar
k
ed
b
y
th
e
r
ed
r
ec
tan
g
le,
was u
tili
ze
d
f
o
r
s
u
b
s
eq
u
en
t a
n
aly
s
is
an
d
p
r
o
ce
s
s
in
g
in
th
is
s
tu
d
y
.
Fig
u
r
e
5
.
A
d
em
o
n
s
tr
atio
n
o
f
a
cu
s
to
m
o
b
ject
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
Th
e
r
o
a
d
co
n
d
itio
n
s
d
etec
tio
n
u
s
in
g
th
e
co
n
vo
lu
tio
n
a
l n
eu
r
a
l
n
etw
o
r
k
(
S
u
jitt
r
a
S
a
-
n
g
iem
)
335
Fig
u
r
e
6
.
T
h
e
ex
am
p
le
o
f
la
n
e
d
etec
tio
n
Fig
u
r
e
7
.
T
h
e
im
ag
e
a
r
ea
3
.
3
.3
.
O
bje
ct
l
a
bellin
g
T
h
e
f
r
a
m
e
ex
tr
ac
tio
n
p
r
o
ce
d
u
r
e
was
en
h
an
ce
d
f
o
llo
win
g
th
e
co
llectio
n
o
f
th
e
v
id
eo
d
ata.
T
h
e
s
u
b
s
eq
u
en
t
p
h
ase
en
tails
o
b
ject
lab
elin
g
,
ex
ec
u
ted
with
L
ab
elM
e,
L
ab
elI
m
g
,
an
d
L
a
b
er
u
i
n
th
is
s
tu
d
y
.
T
h
ese
to
o
ls
ex
h
ib
it
v
a
r
y
in
g
u
s
ag
e
m
eth
o
d
o
lo
g
ies,
with
L
ab
er
u
b
ei
n
g
n
o
tab
ly
u
s
er
-
f
r
ien
d
ly
a
n
d
a
b
le
to
s
u
p
p
o
r
t
m
a
n
y
u
s
er
s
co
n
cu
r
r
e
n
tly
.
T
h
e
l
a
b
elin
g
f
in
d
in
g
s
f
r
o
m
L
ab
er
u
ar
e
r
ec
o
r
d
ed
i
n
.
x
m
l
f
o
r
m
at,
w
h
ich
ca
n
b
e
im
m
e
d
iately
em
p
lo
y
ed
in
later
s
tag
es.
Fo
r
i
n
s
tan
ce
,
Fig
u
r
es
8
to
1
0
illu
s
tr
ate
th
e
an
n
o
tated
in
s
tan
ce
s
o
f
th
e
th
r
ee
o
b
jects.
T
h
e
d
ata
f
o
r
th
is
in
v
esti
g
atio
n
is
g
ath
er
ed
u
tili
zin
g
s
m
ar
tp
h
o
n
es
m
o
u
n
ted
o
n
th
e
ca
r
'
s
win
d
s
h
ield
.
T
h
u
s
,
th
e
d
ata
co
llectio
n
is
co
n
f
in
ed
to
th
e
r
eg
io
n
r
ig
h
t
ah
ea
d
o
f
th
e
v
e
h
icle.
T
h
is
r
esear
ch
p
r
im
ar
ily
f
o
c
u
s
es
o
n
o
b
jects
s
itu
ated
f
o
u
r
m
eter
s
f
r
o
m
th
e
v
eh
icle.
T
h
e
item
s
ar
e
d
esig
n
ate
d
with
r
ec
tan
g
les
th
at
m
u
s
t
co
n
f
o
r
m
to
th
e
f
r
am
e
.
I
n
t
h
e
co
n
te
x
t o
f
ad
jace
n
t f
r
am
es,
th
e
r
ec
tan
g
le
s
o
u
g
h
t to
i
n
ter
s
ec
t a
s
litt
le
as
f
ea
s
ib
le.
As
illu
s
tr
ated
in
Fig
u
r
es
8
-
1
0
,
ea
ch
o
b
ject
h
as
m
u
ltip
le
d
if
f
er
en
t
ch
a
r
ac
ter
is
tics
.
A
m
an
h
o
le
co
v
er
ca
n
p
o
s
s
ess
b
o
th
cir
cu
lar
an
d
r
ec
tan
g
u
lar
f
o
r
m
s
.
T
ab
le
2
s
u
m
m
ar
izes
th
e
o
b
ject
lab
elin
g
r
esu
lts
,
r
ev
ea
lin
g
a
to
tal
o
f
1
7
,
3
9
1
p
h
o
t
o
s
o
f
cr
ac
k
s
,
1
4
,
6
0
6
im
ag
es
o
f
m
an
h
o
les,
an
d
3
,
5
7
3
im
ag
es
o
f
p
o
th
o
les
in
s
eq
u
en
tial
s
eq
u
en
ce
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
40
,
No
.
1
,
Octo
b
er
20
25
:
327
-
3
4
5
336
Fig
u
r
e
8.
I
ll
u
s
tr
atio
n
o
f
a
lab
el
ed
cr
ac
k
Fig
u
r
e
9.
I
ll
u
s
tr
atio
n
of
a
la
b
e
led
p
o
th
o
le
Fig
u
r
e
1
0
.
I
llu
s
tr
atio
n
of
a
lab
eled
m
an
h
o
le
T
ab
le
2
.
T
h
e
co
u
n
t o
f
o
b
ject
l
ab
elin
g
C
r
a
c
k
P
o
t
h
o
l
e
M
a
n
h
o
l
e
1
7
,
3
9
1
3
,
5
7
3
1
4
,
6
0
6
3
.
4
.
Co
nv
o
lutio
na
l
n
eura
l
n
et
wo
rk
T
h
e
co
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
et
wo
r
k
(
C
NN)
o
r
ig
in
ated
in
th
e
1
9
8
0
s
,
in
v
en
ted
b
y
Yan
n
L
eCu
n
.
T
h
e
f
ir
s
t
ar
ch
itectu
r
e
is
L
eNe
t,
wh
ich
was
u
tili
ze
d
f
o
r
d
ig
it
al
r
ec
o
g
n
itio
n
jo
b
s
.
T
h
e
ar
c
h
itectu
r
e
co
m
p
r
is
es
co
n
v
o
l
u
tio
n
al
lay
er
s
,
p
o
o
lin
g
lay
er
s
,
an
d
a
f
u
lly
lin
k
ed
la
y
er
.
T
h
e
s
u
b
s
eq
u
en
t
ar
ch
itect
u
r
es
ar
e
Alex
Net,
I
m
ag
eNe
t,
VGGN
et,
Go
o
g
le
Net,
an
d
R
esNet,
in
th
at
o
r
d
e
r
.
I
n
o
b
ject
d
etec
tio
n
,
th
e
C
NN
ca
n
id
en
tif
y
an
d
lo
ca
te
s
ev
er
al
th
in
g
s
in
s
id
e
an
im
ag
e,
a
task
th
at
is
m
o
r
e
co
m
p
lex
th
a
n
class
if
icatio
n
task
s
.
YOL
O,
an
ac
r
o
n
y
m
f
o
r
"Y
o
u
On
ly
L
o
o
k
On
ce
,
"
is
a
wid
ely
u
tili
ze
d
o
b
j
ec
t
id
en
tific
atio
n
tech
n
iq
u
e
d
e
v
elo
p
ed
b
y
J
o
s
ep
h
R
ed
m
o
n
[
3
8
]
.
T
h
is
ca
n
ef
f
ic
ien
tly
d
etec
t
m
an
y
item
s
.
YOL
O
p
er
f
o
r
m
s
th
is
b
y
u
tili
zin
g
a
co
m
p
le
x
g
r
i
d
f
r
am
ewo
r
k
in
ea
c
h
lay
er
.
T
h
e
tech
n
iq
u
e
i
n
v
o
lv
es
p
ar
titi
o
n
i
n
g
th
e
im
a
g
e
in
to
a
n
ar
r
o
w
win
d
o
w
o
f
s
ize
N*
N
an
d
em
p
lo
y
in
g
an
alg
o
r
ith
m
t
o
an
ticip
ate
t
h
e
o
b
ject.
T
h
ese
p
r
o
ce
d
u
r
es
u
tili
ze
d
ee
p
lear
n
in
g
.
YOL
O
V2
(
o
r
YOL
O9
0
0
0
)
was
cr
ea
ted
to
im
p
r
o
v
e
o
n
YOL
O
V1
,
wh
ich
id
en
tifie
d
o
b
jects
in
r
ea
l
-
tim
e.
Dar
k
n
et1
9
s
er
v
es
as th
e
f
o
u
n
d
atio
n
f
o
r
YOL
O
V2
[
3
9
]
.
YOL
O
V3
h
as d
ev
elo
p
ed
f
r
o
m
v
ar
io
u
s
ar
ch
itectu
r
es,
in
clu
d
in
g
R
esNet
an
d
f
ea
tu
r
e
-
p
y
r
a
m
id
n
etwo
r
k
(
FP
N)
.
T
h
e
f
ea
tu
r
e
ex
tr
ac
tio
n
tech
n
iq
u
e
in
YOL
O
V3
u
tili
ze
s
Dar
k
n
et5
3
,
a
d
ee
p
n
e
u
r
al
n
etwo
r
k
co
m
p
r
is
in
g
5
3
lay
e
r
s
.
I
n
itially
,
Dar
k
n
et
5
3
was
tr
ain
ed
u
s
in
g
th
e
I
m
a
g
en
et
d
ataset.
T
h
e
1
0
6
lay
er
s
an
d
f
ea
tu
r
es
wer
e
Evaluation Warning : The document was created with Spire.PDF for Python.