I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
5
,
Octo
b
er
20
25
,
p
p
.
4
9
4
2
~
4
9
5
3
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
15
i
5
.
pp
4
9
4
2
-
4
9
5
3
4942
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Rea
l t
ime
objec
t
detec
tion for a
dv
a
nced driver a
ss
is
tance
sy
stems us
ing
dee
p learning
t
echni
ques
Su
da
rsh
a
n Siv
a
k
um
a
r,
Sh
ikh
a
T
ripa
t
hi
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s
a
n
d
C
o
mm
u
n
i
c
a
t
i
o
n
,
P
ES U
n
i
v
e
r
si
t
y
,
B
a
n
g
a
l
o
r
e
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
y
2
5
,
2
0
2
5
R
ev
is
ed
J
u
l 3
,
2
0
2
5
Acc
ep
ted
J
u
l 1
2
,
2
0
2
5
Ob
jec
t
d
e
tec
ti
o
n
p
la
y
s
a
c
rit
ica
l
ro
le
in
a
d
v
a
n
c
e
d
d
r
iv
e
r
a
ss
istan
c
e
sy
ste
m
s
(AD
AS),
wh
e
re
ti
m
e
ly
a
n
d
a
c
c
u
ra
te
d
e
tec
ti
o
n
o
f
o
b
jec
ts
o
n
ro
a
d
i
s
e
ss
e
n
ti
a
l
fo
r
v
e
h
icu
lar
sa
fe
ty
.
In
th
is
stu
d
y
,
we
p
ro
p
o
se
a
n
d
e
v
a
lu
a
te
d
e
e
p
lea
rn
in
g
-
b
a
se
d
o
b
jec
t
d
e
tec
ti
o
n
tec
h
n
i
q
u
e
s
—
sp
e
c
ifi
c
a
ll
y
,
c
o
n
v
o
lu
ti
o
n
a
l
n
e
u
ra
l
n
e
two
rk
s
(CNN
)
a
n
d
d
e
n
se
n
e
u
r
a
l
n
e
two
r
k
s
fo
r
re
a
l
-
ti
m
e
o
b
jec
t
d
e
tec
ti
o
n
.
Th
e
p
ro
p
o
se
d
m
o
d
e
l
is
train
e
d
o
n
a
p
u
b
li
c
ly
a
v
a
il
a
b
le
ima
g
e
d
a
tas
e
t
d
e
m
o
n
stra
ti
n
g
it
s
p
o
ten
ti
a
l
t
o
e
n
h
a
n
c
e
th
e
re
li
a
b
il
it
y
o
f
AD
AS
s
y
ste
m
s
with
o
u
t
th
e
u
se
o
f
a
n
ima
g
e
p
re
p
ro
c
e
ss
in
g
b
lo
c
k
.
He
re
th
e
sy
ste
m
a
u
to
m
a
ti
c
a
ll
y
sto
p
s
with
o
u
t
a
n
y
h
u
m
a
n
i
n
terv
e
n
t
io
n
.
Ou
r
re
su
lt
s
h
ig
h
li
g
h
t
th
e
stre
n
g
th
s
a
n
d
l
imitatio
n
s
o
f
u
sin
g
CIF
AR
-
1
0
,
CIF
AR
-
1
0
0
a
n
d
YO
LO
d
a
tas
e
ts
fo
r
tran
sfe
r
lea
rn
in
g
,
p
re
-
train
in
g
a
n
d
a
lg
o
rit
h
m
c
las
sifica
ti
o
n
.
Im
p
ro
v
e
m
e
n
ts
i
n
m
o
d
e
l
o
p
ti
m
iza
ti
o
n
a
n
d
h
a
rd
wa
re
in
teg
ra
ti
o
n
h
a
v
e
b
e
e
n
a
c
h
iev
e
d
u
sin
g
h
a
rd
wa
re
in
l
o
o
p
(HIL)
se
t
u
p
.
Th
e
m
o
d
e
ls
a
re
e
v
a
lu
a
ted
o
n
CIF
AR
-
1
0
,
CIF
AR
-
1
0
0
a
n
d
YO
LO
d
a
tas
e
ts,
with
a
f
o
c
u
s
o
n
th
e
imp
a
c
t
o
f
ima
g
e
p
re
-
p
ro
c
e
ss
in
g
o
n
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
a
n
d
sp
e
e
d
.
Ex
p
e
rime
n
tal
re
su
lt
s
sh
o
w
th
a
t
th
e
p
ro
p
o
se
d
a
l
g
o
rit
h
m
o
u
tp
e
rf
o
rm
s
th
e
p
re
v
io
u
s
m
e
th
o
d
s,
b
y
a
c
h
iev
in
g
a
b
e
tt
e
r
a
c
c
u
ra
c
y
,
c
o
n
t
rib
u
ti
n
g
to
sa
fe
r
a
n
d
ro
b
u
st
sy
ste
m
with
o
u
t
a
n
a
d
d
it
io
n
a
l
ima
g
e
p
re
p
r
o
c
e
ss
in
g
b
l
o
c
k
.
K
ey
w
o
r
d
s
:
Ad
v
an
ce
d
d
r
iv
er
ass
is
tan
ce
s
y
s
tem
s
An
ti
-
lo
ck
b
r
a
k
in
g
s
y
s
tem
AR
S 6
2
0
R
ADA
R
C
o
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
elec
tr
o
n
ic
co
n
tr
o
l u
n
it
Har
d
war
e
in
lo
o
p
SVC
3
0
0
C
am
er
a
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
:
Su
d
ar
s
h
an
Siv
ak
u
m
ar
Dep
ar
tm
en
t o
f
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
icatio
n
,
PES U
n
iv
e
r
s
ity
1
0
0
f
ee
t
r
in
g
r
o
ad
,
B
an
ash
an
k
ar
i stag
e
3
,
Dwa
r
ak
a
Nag
a
r
,
B
an
g
alo
r
e,
Ka
r
n
atak
a
5
6
0
0
8
5
,
I
n
d
ia
E
m
ail:
s
u
d
ar
s
h
an
s
iv
ak
u
m
a
r
9
9
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
R
o
ad
s
af
ety
r
em
ain
s
a
g
lo
b
al
co
n
ce
r
n
,
with
tr
af
f
ic
-
r
elate
d
in
cid
en
ts
r
esu
ltin
g
in
s
ig
n
if
ican
t
lo
s
s
o
f
life
an
d
p
r
o
p
e
r
ty
.
I
n
r
esp
o
n
s
e,
ad
v
an
ce
d
d
r
iv
er
ass
is
tan
ce
s
y
s
tem
s
(
ADAS)
h
av
e
em
er
g
ed
as
a
c
r
itical
co
m
p
o
n
en
t
o
f
m
o
d
er
n
v
eh
icl
es,
aim
in
g
to
r
ed
u
ce
h
u
m
a
n
er
r
o
r
an
d
im
p
r
o
v
e
d
r
iv
in
g
o
u
tco
m
es
th
r
o
u
g
h
th
e
in
teg
r
atio
n
o
f
s
en
s
in
g
an
d
d
ec
is
io
n
-
m
ak
in
g
tech
n
o
lo
g
ies.
A
m
o
n
g
th
ese,
o
b
ject
d
etec
tio
n
p
lay
s
a
v
ital
r
o
le
i
n
r
ec
o
g
n
izin
g
p
ed
estrian
s
,
v
e
h
ic
les,
tr
af
f
ic
s
ig
n
s
,
an
d
o
b
s
tacle
s
,
en
ab
lin
g
tim
ely
in
ter
v
e
n
tio
n
s
s
u
ch
as a
u
to
m
atic
b
r
ak
in
g
o
r
s
teer
in
g
co
r
r
ec
tio
n
s
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
d
e
ep
lear
n
in
g
h
av
e
en
a
b
led
s
u
b
s
tan
tial
p
r
o
g
r
ess
in
co
m
p
u
ter
v
is
io
n
task
s
,
in
clu
d
i
n
g
o
b
ject
d
etec
tio
n
an
d
class
if
icatio
n
.
C
o
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs)
h
av
e
p
r
o
v
en
h
ig
h
l
y
ef
f
ec
tiv
e
i
n
ex
tr
ac
tin
g
s
p
atial
f
ea
tu
r
es
f
r
o
m
v
is
u
al
d
ata,
wh
ile
r
ec
u
r
r
e
n
t
n
eu
r
al
n
etwo
r
k
s
(
R
NNs)
o
f
f
er
tem
p
o
r
al
m
o
d
elin
g
ca
p
ab
ilit
ies,
b
en
ef
icial
f
o
r
d
y
n
am
ic
d
r
iv
in
g
en
v
ir
o
n
m
en
ts
.
T
h
is
p
ap
er
p
r
o
p
o
s
es
a
C
NN
-
R
NN
h
y
b
r
id
m
o
d
el
f
o
r
o
b
ject
d
etec
tio
n
,
with
a
f
o
cu
s
o
n
its
im
p
lem
en
tati
o
n
in
an
em
b
e
d
d
ed
au
to
m
o
tiv
e
e
n
v
ir
o
n
m
en
t.
Ob
ject
d
etec
tio
n
is
a
f
u
n
d
a
m
en
tal
task
in
co
m
p
u
ter
v
is
io
n
th
at
in
v
o
lv
es
id
en
tify
i
n
g
a
n
d
lo
ca
lizin
g
o
b
jects
with
in
im
ag
es
o
r
v
id
eo
f
r
am
es.
W
ith
r
ap
id
ad
v
an
ce
m
en
ts
in
d
ee
p
lear
n
in
g
,
o
b
ject
d
etec
tio
n
tech
n
iq
u
es
h
av
e
ac
h
iev
e
d
r
em
ar
k
ab
le
ac
cu
r
ac
y
an
d
ar
e
n
o
w
wid
ely
d
ep
lo
y
e
d
in
v
ar
io
u
s
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
,
in
clu
d
in
g
s
ec
u
r
it
y
s
u
r
v
eillan
ce
,
r
o
b
o
tics
,
h
ea
lt
h
ca
r
e,
an
d
in
tellig
en
t
tr
an
s
p
o
r
tatio
n
s
y
s
tem
s
.
On
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l time
o
b
ject
d
etec
tio
n
fo
r
a
d
va
n
ce
d
d
r
iver a
s
s
is
ta
n
ce
s
ystems
…
(
S
u
d
a
r
s
h
a
n
S
iva
k
u
m
a
r
)
4943
o
f
th
e
m
o
s
t
cr
itical
ap
p
licatio
n
s
o
f
o
b
ject
d
etec
tio
n
is
in
A
DAS,
wh
er
e
th
e
ab
ilit
y
t
o
d
e
tect
an
d
r
ec
o
g
n
ize
v
eh
icles,
p
ed
estrian
s
,
tr
a
f
f
ic
s
ig
n
s
,
an
d
o
th
er
r
o
a
d
elem
e
n
ts
in
r
ea
l
-
tim
e
is
ess
en
tial
f
o
r
e
n
s
u
r
in
g
d
r
iv
er
an
d
p
ass
en
g
er
s
af
ety
.
T
h
ese
s
y
s
tem
s
r
ely
o
n
ac
cu
r
ate
v
is
u
al
p
er
ce
p
tio
n
t
o
ass
is
t
with
task
s
s
u
ch
as
lan
e
d
e
p
ar
tu
r
e
war
n
in
g
s
,
co
llis
io
n
av
o
id
an
ce
,
an
d
au
t
o
m
ated
b
r
ak
in
g
.
R
ap
id
ad
v
an
ce
m
e
n
t
in
a
u
to
m
o
tiv
e
tech
n
o
l
o
g
y
h
as
led
to
th
e
d
ev
elo
p
m
en
t
o
f
ADAS,
wh
ich
aim
s
to
r
ed
u
ce
h
u
m
a
n
er
r
o
r
s
a
n
d
en
h
an
ce
d
r
i
v
in
g
s
af
ety
.
O
b
ject
d
etec
tio
n
is
a
f
u
n
d
a
m
en
tal
c
o
m
p
o
n
e
n
t
o
f
ADAS,
en
ab
lin
g
v
e
h
icles
to
id
en
tify
p
ed
estrian
s
,
o
th
er
v
eh
icles,
tr
af
f
ic
s
ig
n
s
,
an
d
o
b
s
tacle
s
o
n
th
e
r
o
ad
.
Var
io
u
s
d
etec
tio
n
tech
n
iq
u
es,
in
clu
d
in
g
im
ag
e
p
r
o
ce
s
s
in
g
,
m
ac
h
in
e
lear
n
in
g
,
a
n
d
d
ee
p
lear
n
in
g
,
h
av
e
b
ee
n
a
p
p
lied
t
o
im
p
r
o
v
e
ac
c
u
r
ac
y
an
d
e
f
f
icien
cy
[
1
]
,
[
2
]
.
T
r
ad
itio
n
al
o
b
ject
d
etec
tio
n
t
ec
h
n
iq
u
es
r
ely
o
n
im
ag
e
p
r
o
ce
s
s
in
g
an
d
f
ea
tu
r
e
ex
tr
ac
tio
n
m
eth
o
d
s
s
u
ch
as
ed
g
e
d
etec
tio
n
,
co
l
o
r
s
eg
m
en
tatio
n
,
an
d
tem
p
late
m
atch
in
g
.
T
h
ese
m
eth
o
d
s
o
f
ten
r
eq
u
ir
e
h
a
n
d
cr
af
te
d
f
ea
tu
r
es
an
d
a
r
e
s
en
s
itiv
e
to
ch
an
g
es
in
lig
h
tin
g
an
d
e
n
v
ir
o
n
m
en
tal
co
n
d
itio
n
s
[
3
]
.
Ma
ch
in
e
lear
n
in
g
alg
o
r
ith
m
s
,
s
u
ch
as
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
an
d
r
a
n
d
o
m
f
o
r
ests
(
R
Fs
)
,
im
p
r
o
v
e
o
b
ject
d
etec
tio
n
b
y
lear
n
in
g
p
atter
n
s
f
r
o
m
lab
eled
d
ata.
Ho
wev
er
,
th
ese
m
eth
o
d
s
s
till
r
ely
o
n
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
a
r
e
lim
ited
in
th
eir
ab
ilit
y
to
h
an
d
le
c
o
m
p
le
x
en
v
ir
o
n
m
en
ts
[
4
]
.
Dee
p
lear
n
in
g
h
as
r
ev
o
lu
tio
n
ized
o
b
ject
d
etec
tio
n
in
ADAS
b
y
lev
er
ag
in
g
C
NNs a
n
d
o
th
er
n
eu
r
al
n
etwo
r
k
ar
ch
itectu
r
es.
Po
p
u
lar
m
o
d
els s
u
ch
as Fast
er
R
-
C
N
N,
y
o
u
o
n
ly
lo
o
k
o
n
ce
(
YOL
O
)
,
an
d
s
in
g
l
e
s
h
o
t
m
u
ltib
o
x
d
etec
to
r
(
SS
D)
h
a
v
e
d
em
o
n
s
tr
ated
h
ig
h
ac
c
u
r
ac
y
an
d
r
ea
l
-
tim
e
p
er
f
o
r
m
an
ce
[
5
]
.
T
h
ese
m
o
d
els
au
to
m
atica
lly
lear
n
h
ier
ar
ch
ical
f
ea
tu
r
es
an
d
ca
n
a
d
ap
t
to
v
ar
io
u
s
d
r
iv
i
n
g
co
n
d
itio
n
s
[
6
]
,
[
7
]
.
T
h
er
e
is
in
cr
ea
s
in
g
d
em
an
d
b
y
cu
s
to
m
er
s
f
o
r
en
h
a
n
ce
d
ADAS
ex
p
er
ien
ce
with
im
p
r
o
v
ed
s
af
ety
to
d
r
iv
er
a
n
d
p
e
d
estrian
s
.
I
n
th
e
y
ea
r
2
0
2
2
t
h
er
e
wer
e
o
n
ly
2
o
r
ig
in
al
eq
u
ip
m
en
t
m
a
n
u
f
ac
t
u
r
er
s
an
d
in
t
h
e
y
ea
r
2
0
2
3
th
er
e
wer
e
7
OE
M’
s
th
a
t
ad
ap
ted
th
e
ADAS
s
y
s
tem
s
to
th
eir
ca
r
s
.
c
o
m
p
ar
e
d
to
m
a
n
y
o
th
e
r
co
u
n
tr
ies,
I
n
d
ia
is
h
av
in
g
m
o
s
t
co
m
p
l
ex
u
s
e
ca
s
es
d
u
e
to
it
s
p
o
p
u
latio
n
d
en
s
ity
an
d
r
o
a
d
co
n
d
itio
n
s
.
T
o
d
ay
th
e
ad
v
an
ce
m
e
n
t o
f
Ad
v
an
ce
d
d
r
i
v
er
ass
is
tan
t sy
s
tem
s
h
as r
esu
l
ted
in
in
cr
ea
s
ed
f
u
el
ef
f
icien
c
y
.
As
p
er
a
s
tu
d
y
b
y
Glo
b
al
E
d
g
e
So
f
t,
th
e
f
u
tu
r
e
ad
v
a
n
ce
d
d
r
iv
er
ass
is
tan
ce
s
y
s
tem
s
wo
u
ld
h
av
e
wir
eless
n
etwo
r
k
co
n
n
ec
tiv
it
y
th
at
ca
n
ea
s
ily
b
e
in
s
talled
in
ca
r
s
.
I
n
o
th
e
r
wo
r
d
s
,
ca
r
s
will
co
m
m
u
n
icate
b
etter
,
r
esu
ltin
g
in
a
s
af
er
an
d
m
o
r
e
co
n
v
e
n
ien
t
au
to
m
ated
d
r
iv
in
g
ex
p
e
r
ien
ce
.
I
t
h
as
b
ee
n
p
r
ed
icted
b
y
m
a
n
y
au
to
m
o
tiv
e
OE
M’
s
t
h
at
ADAS
is
g
o
in
g
t
o
d
o
m
in
ate
I
n
d
ia
n
m
ar
k
ets
in
an
o
th
er
3
-
4
y
ea
r
s
ev
en
in
th
e
m
id
v
ar
ian
t
ca
r
s
.
B
y
th
e
y
ea
r
2
0
2
7
it’s
ex
p
ec
ted
t
h
at
L
4
(
d
r
iv
e
r
less
)
ADAS
s
y
s
tem
s
will
b
e
av
ailab
le
i
n
m
a
n
y
co
u
n
tr
ies.
ADAS
ar
e
cr
itical
in
r
ed
u
cin
g
r
o
ad
ac
cid
en
ts
b
y
m
in
im
izi
n
g
h
u
m
a
n
er
r
o
r
.
T
h
ese
s
y
s
tem
s
p
r
o
v
id
e
f
u
n
ctio
n
alities
s
u
ch
as
p
ed
es
tr
ian
d
etec
tio
n
,
lan
e
d
ep
ar
t
u
r
e
war
n
in
g
,
au
t
o
m
atic
em
er
g
en
cy
b
r
ak
in
g
,
a
n
d
co
llis
io
n
av
o
id
an
ce
.
Ho
wev
e
r
,
co
n
v
e
n
tio
n
al
ad
v
a
n
ce
d
d
r
iv
er
ass
is
tan
ce
im
p
lem
en
tatio
n
s
s
tr
u
g
g
le
in
lo
w
-
v
is
ib
ilit
y
co
n
d
itio
n
s
,
af
f
ec
ti
n
g
d
etec
tio
n
ac
c
u
r
ac
y
an
d
in
cr
ea
s
in
g
ac
cid
en
t
r
is
k
s
.
Desp
ite
s
ig
n
if
ican
t
ad
v
an
ce
m
e
n
ts
,
o
b
ject
d
etec
tio
n
in
ADAS
f
ac
es
s
ev
er
al
c
h
allen
g
es,
in
clu
d
in
g
r
ea
l
-
tim
e
Pr
o
ce
s
s
in
g
,
en
s
u
r
in
g
f
ast
an
d
ef
f
icien
t
d
etec
tio
n
wh
ile
m
ain
tain
in
g
ac
cu
r
ac
y
[
8
]
.
Ad
v
er
s
e
wea
th
er
co
n
d
itio
n
s
h
an
d
lin
g
v
ar
iatio
n
s
in
lig
h
tin
g
,
f
o
g
,
r
ain
,
an
d
s
n
o
w
is
th
e
s
ec
o
n
d
c
h
allen
g
e
[
9
]
.
Occ
lu
s
io
n
s
an
d
d
y
n
am
ic
en
v
ir
o
n
m
e
n
ts
,
wh
er
e
o
b
jects a
r
e
p
ar
tially
h
i
d
d
en
o
r
m
o
v
in
g
u
n
p
r
ed
ictab
ly
is
also
a
ch
allen
g
e
[
1
0
]
.
Un
lik
e
tr
ad
itio
n
al
ap
p
r
o
ac
h
es
th
at
r
eq
u
ir
e
s
ig
n
if
ican
t
p
r
e
p
r
o
ce
s
s
in
g
o
r
h
ig
h
co
m
p
u
tatio
n
al
p
o
wer
,
th
e
p
r
o
p
o
s
ed
m
o
d
el
in
th
is
p
a
p
er
ac
h
iev
es
co
m
p
etitiv
e
ac
cu
r
ac
y
u
s
in
g
r
aw
im
ag
es
f
r
o
m
t
h
e
C
I
FAR
-
1
0
an
d
C
I
FAR
-
1
0
0
d
atasets
.
Fu
r
th
er
m
o
r
e,
we
v
alid
ate
t
h
e
r
ea
l
-
tim
e
ap
p
licab
ilit
y
o
f
t
h
e
s
y
s
tem
b
y
in
teg
r
atin
g
it
with
an
an
ti
-
lo
ck
b
r
ak
in
g
s
y
s
tem
(
AB
S)
elec
tr
o
n
ic
co
n
tr
o
l u
n
it (
E
C
U)
an
d
s
im
u
latin
g
r
esp
o
n
s
e
s
th
r
o
u
g
h
h
ar
d
war
e
-
in
-
th
e
-
lo
o
p
(
H
I
L
)
test
in
g
.
T
h
is
r
esear
ch
b
r
i
d
g
es
th
e
g
a
p
b
etwe
en
alg
o
r
ith
m
ic
d
ev
elo
p
m
en
t
an
d
p
r
ac
tical
d
ep
lo
y
m
e
n
t
in
v
eh
ic
u
lar
s
y
s
tem
s
.
Fu
tu
r
e
r
esear
ch
aim
s
to
e
n
h
an
ce
m
o
d
el
r
o
b
u
s
tn
ess
,
in
teg
r
ate
s
en
s
o
r
f
u
s
io
n
tech
n
iq
u
es
(
s
u
ch
as
L
iDAR
an
d
r
a
d
ar
)
,
an
d
d
ev
elo
p
en
e
r
g
y
-
ef
f
icien
t
alg
o
r
ith
m
s
f
o
r
d
e
p
lo
y
m
en
t
in
em
b
ed
d
ed
au
to
m
o
tiv
e
s
y
s
tem
s
[
1
1
]
–
[
1
3
]
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
Fu
zz
y
lo
g
ic
co
n
tr
o
lled
a
n
tilo
ck
b
r
ak
i
n
g
s
y
s
tem
was
b
asically
in
tr
o
d
u
ce
d
in
th
e
AB
S
s
y
s
tem
s
f
o
r
im
p
r
o
v
in
g
th
e
b
id
ir
ec
tio
n
al
s
tab
ilit
y
o
v
er
th
e
u
n
if
o
r
m
an
d
n
o
n
-
u
n
if
o
r
m
r
o
ad
s
u
r
f
ac
es.
T
h
e
p
er
f
o
r
m
a
n
ce
f
o
r
f
ew
u
s
e
ca
s
es
was
n
o
t
s
atis
f
ac
to
r
y
.
Fo
r
ex
am
p
le,
th
e
d
r
i
v
er
m
u
s
t
tak
e
a
n
ac
tio
n
af
ter
d
et
ec
tin
g
th
e
o
b
s
tacle
.
T
h
e
s
y
s
tem
f
ailed
wh
en
th
e
o
b
s
tacle
s
wer
e
to
o
cl
o
s
e
as
th
er
e
was
n
o
m
ec
h
a
n
is
m
in
t
h
e
s
y
s
tem
to
in
d
icate
th
e
d
r
iv
er
wh
e
n
ev
er
t
h
e
o
b
s
tacle
i
s
f
ew
m
eter
s
ah
ea
d
o
f
th
e
v
e
h
icle
[
1
]
.
T
h
e
wo
r
k
b
y
L
iu
et
a
l.
u
s
es
a
co
m
b
in
atio
n
o
f
elec
tr
o
m
ec
h
a
n
ical
b
r
ak
i
n
g
s
y
s
tem
s
an
d
f
ea
tu
r
e
b
ased
alg
o
r
ith
m
th
at
h
elp
s
to
s
h
o
r
te
n
th
e
b
r
a
k
in
g
d
is
tan
ce
th
er
eb
y
r
esu
ltin
g
in
im
p
r
o
v
e
d
b
r
a
k
in
g
s
tab
ilit
y
co
m
p
ar
ed
to
p
r
ev
io
u
s
s
y
s
tem
.
I
t
u
s
es
el
ec
tr
ic
en
er
g
y
as
th
e
en
er
g
y
s
o
u
r
ce
o
f
th
e
b
r
ak
in
g
s
y
s
tem
an
d
u
s
es
a
m
o
to
r
to
d
r
iv
e
th
e
ac
tu
ato
r
to
g
en
er
ate
b
r
ak
i
n
g
f
o
r
ce
,
wh
ich
r
esp
o
n
d
s
q
u
ick
ly
.
T
h
is
ac
tio
n
m
a
y
a
f
f
ec
t
th
e
life
tim
e
o
f
th
e
s
y
s
tem
.
Ag
ain
,
th
er
e
is
n
o
m
ec
h
an
is
m
s
in
th
e
s
y
s
tem
to
h
alt
it
o
r
aler
t
th
e
d
r
iv
er
wh
en
an
o
b
s
tacle
is
d
etec
ted
[
2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
4
2
-
4
9
5
3
4944
W
ith
th
e
in
tr
o
d
u
ctio
n
o
f
f
u
zz
y
PID
co
n
tr
o
ller
s
alo
n
g
with
t
h
e
co
n
t
r
o
l
alg
o
r
ith
m
in
th
e
A
B
S
s
y
s
tem
s
p
r
o
m
is
in
g
r
esu
lts
wer
e
ac
h
iev
ed
co
m
p
ar
e
d
to
th
e
p
r
e
v
io
u
s
wo
r
k
s
.
I
t
r
ed
u
ce
d
th
e
b
r
ak
i
n
g
d
is
tan
ce
b
y
u
p
to
1
0
%
co
m
p
ar
ed
to
a
co
n
v
en
t
io
n
al
PID
an
d
b
y
m
o
r
e
t
h
an
3
0
%
co
m
p
ar
e
d
to
a
r
ela
y
co
n
tr
o
ller
.
Gr
ad
u
al
im
p
r
o
v
em
e
n
ts
to
war
d
s
s
to
p
p
a
g
e
o
f
v
eh
icles
wer
e
s
ee
n
in
co
n
ju
n
ctio
n
with
th
e
AB
S
s
y
s
te
m
s
in
th
e
u
s
e
ca
s
es
wh
er
e
th
e
o
b
s
tacle
s
wer
e
cl
o
s
er
to
th
e
o
b
ject.
B
u
t
th
ese
s
y
s
tem
s
lack
ed
to
ca
lcu
late
th
e
d
is
tan
ce
o
f
o
b
s
tacle
s
wh
ich
is
a
k
ey
f
ac
to
r
an
d
p
r
im
ar
ily
f
o
cu
s
ed
o
n
r
e
d
u
ctio
n
o
f
b
r
ak
in
g
d
is
tan
ce
[
3
]
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
b
y
Ur
m
at
et
a
l.
u
s
es
3
2
-
b
it
AR
M
b
ased
em
b
ed
d
e
d
s
y
s
tem
s
with
ac
ce
ler
o
m
eter
,
g
y
r
o
s
co
p
e
an
d
m
ag
n
et
o
m
eter
.
GNSS
m
o
d
u
le
alo
n
g
with
B
L
E
was
u
s
ed
to
d
etec
t
th
e
p
ed
estrian
s
b
y
s
im
u
latin
g
p
e
d
estrian
’
s
walk
in
g
p
o
s
itio
n
s
in
th
e
r
ea
l
tim
e
s
c
en
ar
io
.
T
h
e
r
esu
lt
s
o
f
ex
p
er
i
m
en
t
ar
e
v
er
if
ied
as
2
.
2
%
d
is
tan
ce
er
r
o
r
,
4
%
m
ax
im
u
m
a
v
er
ag
e
p
o
s
itio
n
in
g
er
r
o
r
an
d
3
.
6
5
%
s
tep
co
u
n
t
er
r
o
r
.
T
h
e
r
esu
lts
wer
e
g
o
o
d
f
o
r
s
h
o
r
t
r
a
n
g
e
d
is
tan
ce
s
b
u
t n
o
t g
o
o
d
f
o
r
l
o
n
g
r
an
g
e
d
is
tan
ce
s
[
4
]
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
in
[
5
]
u
s
es
th
er
m
al
n
ig
h
t
v
is
io
n
s
y
s
te
m
s
to
d
etec
t
th
e
p
ed
estrian
s
i
n
r
ea
l
tim
e
ca
r
d
ash
b
o
ar
d
s
b
y
ex
tr
ac
tin
g
t
h
e
r
eg
i
o
n
o
f
in
ter
ests
(
R
OI
)
th
er
eb
y
ale
r
tin
g
t
h
e
d
r
iv
er
s
.
T
h
e
m
ain
d
r
awb
ac
k
o
f
th
is
m
eth
o
d
is
th
at
s
o
m
e
o
f
t
h
e
p
ed
estrian
s
am
p
les
ar
e
n
o
t
h
ig
h
lig
h
ted
d
u
e
to
th
e
b
ac
k
g
r
o
u
n
d
th
e
r
m
al
en
e
r
g
y
s
u
ch
as
a
ca
r
ex
h
au
s
t,
wh
ich
ca
n
d
is
to
r
t
th
e
ca
n
d
id
ates
s
h
ap
e
an
d
p
u
t
it
o
u
ts
id
e
th
e
s
tr
in
g
en
t
asp
ec
t
r
atio
co
n
s
tr
ain
ts
.
W
ith
th
e
ad
ap
tio
n
o
f
ADAS
s
y
s
tem
s
,
th
e
s
y
s
t
em
in
[
6
]
u
s
es
tech
n
iq
u
es
o
f
im
ag
e
p
r
o
ce
s
s
in
g
,
ad
ap
tiv
e
s
ig
n
al
p
r
o
ce
s
s
in
g
,
d
e
ep
lear
n
in
g
an
d
c
o
m
p
u
ter
v
is
io
n
to
d
etec
t
th
e
r
ea
l
tim
e
p
e
d
estrian
m
o
v
em
en
ts
.
T
h
e
s
y
s
tem
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
7
5
.
8
%.
T
h
e
s
am
p
les
t
ak
en
wer
e
clea
r
p
ed
estrian
s
am
p
les
in
r
ea
l
tim
e.
T
h
e
s
y
s
tem
’
s
b
eh
av
io
r
to
d
e
tect
u
n
d
er
b
ad
wea
th
er
co
n
d
itio
n
s
o
r
wh
en
u
n
clea
r
im
a
g
es
ar
e
f
ed
was
n
o
t
ex
p
er
im
en
ted
.
T
h
e
wo
r
k
in
[
8
]
attem
p
ts
to
aler
t
th
e
p
ed
estrian
s
th
r
o
u
g
h
th
eir
s
m
ar
t
p
h
o
n
es
wh
en
ev
e
r
a
v
eh
icle
co
m
es c
lo
s
e.
A
f
u
lly
f
u
n
ctio
n
i
n
g
m
o
b
ile
s
y
s
tem
h
as b
ee
n
d
e
v
elo
p
ed
wh
e
r
e
p
e
d
estrian
s
g
et
war
n
in
g
s
wh
en
t
h
e
s
m
ar
tp
h
o
n
e
s
en
s
o
r
r
ec
eiv
es
i
n
f
o
r
m
atio
n
o
n
th
e
ar
r
iv
al
o
f
v
eh
icles
th
r
o
u
g
h
d
ir
ec
t
W
i
-
Fi
b
ased
p
ee
r
to
p
ee
r
co
m
m
u
n
icatio
n
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ate
th
at
th
e
en
er
g
y
ef
f
icien
cy
an
d
p
o
s
itio
n
in
g
ac
cu
r
ac
y
o
f
Saf
er
C
r
o
s
s
ar
e
im
p
r
o
v
ed
b
y
5
2
%
an
d
7
2
%
o
n
av
er
ag
e
co
m
p
ar
ed
wi
th
ex
is
tin
g
s
o
lu
tio
n
s
with
m
is
s
in
g
s
u
p
p
o
r
t
f
o
r
p
o
s
itio
n
in
g
ac
cu
r
ac
y
an
d
e
n
e
r
g
y
e
f
f
icien
cy
,
an
d
t
h
e
p
h
o
n
e
-
v
iewin
g
e
v
en
t
d
etec
tio
n
ac
c
u
r
ac
y
is
o
v
er
9
0
%.
T
h
e
av
e
r
ag
e
er
r
o
r
o
f
t
h
e
s
y
s
te
m
is
ar
o
u
n
d
1
.
6
s
ec
s
.
T
h
er
e
is
n
o
s
u
ch
m
ec
h
a
n
is
m
o
f
aler
ti
n
g
th
e
d
r
iv
er
ab
o
u
t
p
ed
estrian
cr
o
s
s
in
g
s
.
T
h
e
s
y
s
te
m
i
n
[
9
]
u
s
es
a
C
NN
b
ase
d
a
p
p
r
o
a
c
h
b
y
a
d
ap
t
in
g
h
ig
h
er
le
v
el
o
f
s
e
m
a
n
ti
cs
,
f
ea
t
u
r
es
,
ar
ch
ite
ct
u
r
e
a
n
d
t
r
a
in
in
g
s
t
r
at
eg
i
es
i
n
o
b
je
ct
d
et
ec
ti
o
n
.
T
h
e
ac
c
u
r
a
cy
a
c
h
ie
v
e
d
b
y
t
ak
in
g
cl
ea
r
s
am
p
l
e
o
f
im
a
g
es
is
a
r
o
u
n
d
6
8
%
.
T
h
e
a
c
cu
r
a
cy
/s
e
n
s
it
iv
it
y
o
f
s
af
et
y
c
r
i
tica
l
s
y
s
t
em
s
s
h
o
u
l
d
r
ea
s
o
n
a
b
l
y
b
e
a
litt
le
h
i
g
h
er
.
T
h
is
s
en
s
iti
v
it
y
/
ac
cu
r
a
c
y
o
f
th
e
s
y
s
te
m
f
o
r
cl
ea
r
p
e
d
estr
ia
n
s
am
p
le
s
e
e
m
s
t
o
b
e
l
o
w.
T
h
e
s
y
s
te
m
i
n
[
1
1
]
u
s
es
s
u
p
e
r
v
is
e
d
,
u
n
s
u
p
er
v
is
ed
,
d
ee
p
a
n
d
r
e
in
f
o
r
ce
m
e
n
t
l
ea
r
n
i
n
g
.
T
h
e
f
o
c
u
s
o
f
t
h
e
ML
r
ese
a
r
ch
i
n
t
h
is
p
a
p
e
r
is
aim
ed
at
r
e
d
u
ci
n
g
th
e
r
o
a
d
t
r
a
f
f
ic
f
at
ali
ties
an
d
a
cc
i
d
en
ts
b
y
u
s
i
n
g
Ha
d
o
o
p
a
n
d
b
i
g
d
at
a
m
e
th
o
d
s
b
y
p
r
e
d
i
cti
n
g
th
e
p
ed
est
r
ia
n
d
e
n
s
it
y
.
I
t
f
o
c
u
s
es
o
n
t
h
e
u
s
e
ca
s
es
s
u
c
h
as
a
u
t
o
m
ati
c
la
n
e
ass
is
t
an
ce
,
a
u
t
o
m
a
tic
p
a
r
k
ass
is
t
an
ce
an
d
s
te
er
in
g
c
o
n
tr
o
l
m
ec
h
an
is
m
s
.
T
h
is
m
ec
h
a
n
is
m
al
er
ts
o
n
l
y
th
e
d
r
i
v
e
r
a
b
o
u
t
t
h
e
a
b
o
v
e
f
a
cto
r
s
.
T
h
e
wo
r
k
in
[
1
2
]
attem
p
ts
t
o
ex
ten
d
th
e
life
tim
e
o
f
AB
S
s
y
s
tem
s
th
at
g
ets
d
eter
io
r
a
ted
d
u
e
to
s
tr
u
ctu
r
al
d
am
a
g
es/wear
an
d
tear
.
I
t
e
x
p
ec
ts
th
e
d
r
iv
er
to
s
ee
th
e
p
e
d
estrian
s
an
d
tak
e
a
ctio
n
.
Fu
zz
y
lo
g
ic
-
b
ased
life
-
ex
ten
d
in
g
co
n
t
r
o
l
(
FLE
C
)
s
y
s
tem
f
o
r
in
cr
ea
s
in
g
th
e
s
er
v
ice
life
o
f
th
e
AB
S is
p
r
o
p
o
s
ed
.
T
h
e
m
ain
co
n
tr
ib
u
tio
n
s
with
th
is
ap
p
r
o
ac
h
as
co
m
p
ar
ed
to
o
th
er
ap
p
r
o
ac
h
es
ar
e
it
d
ec
r
ea
s
ed
s
to
p
p
in
g
d
is
tan
ce
b
y
1
9
.
2
%,
s
to
p
p
in
g
tim
e
b
y
2
0
.
9
%
an
d
wh
ee
l
lo
ck
u
p
tim
e
f
r
o
m
8
3
.
3
%
to
3
.
1
%
o
f
to
tal
b
r
ak
in
g
tim
e.
T
h
er
e
is
n
o
m
ec
h
an
is
m
to
h
alt
th
e
AB
S
b
r
ak
in
g
s
y
s
tem
au
to
m
atica
lly
in
ca
s
e
th
e
o
b
ject
is
v
er
y
clo
s
e
as
th
e
r
ef
lex
ac
tio
n
f
r
o
m
th
e
d
r
iv
e
r
ca
n
n
o
t b
e
e
x
p
ec
ted
in
s
u
ch
ca
s
es.
De
So
u
s
a
et
a
l.
[
1
4
]
u
s
e
NVI
DI
A’
s
em
b
ed
d
ed
c
h
ip
in
t
h
e
ADAS
s
y
s
tem
s
th
at
h
elp
s
th
e
d
r
iv
er
to
m
ak
e
d
ec
is
io
n
s
o
n
d
ay
-
to
-
d
a
y
tr
af
f
ic
s
itu
atio
n
s
.
I
t
u
s
es
a
co
m
b
in
atio
n
o
f
d
ee
p
lear
n
i
n
g
an
d
co
m
p
u
ter
v
is
io
n
tech
n
iq
u
e
f
o
r
la
n
e
s
eg
m
en
tati
o
n
,
d
etec
tin
g
th
e
v
eh
icles/p
ed
estrian
s
in
th
e
r
ea
l
tim
e
tr
af
f
ic
en
v
ir
o
n
m
en
t
a
n
d
aler
ts
th
e
d
r
iv
e
r
ab
o
u
t
c
o
llis
io
n
o
f
th
e
s
y
s
tem
with
ca
r
s
o
r
p
ed
estrian
s
.
T
h
is
u
s
es
R
esNet
ar
ch
itectu
r
e
wh
e
r
e
th
e
ac
cu
r
ac
y
is
6
6
.
6
%.
Ag
ain
,
th
e
s
y
s
tem
’
s
b
eh
av
io
r
f
o
r
u
n
clea
r
s
am
p
les/
d
u
r
in
g
b
a
d
wea
th
er
co
n
d
itio
n
s
ar
e
n
o
t
tak
en
in
to
co
n
s
id
er
atio
n
.
T
h
e
v
eh
icle
in
[
1
5
]
u
s
es
b
ac
k
p
r
o
p
a
g
atio
n
m
eth
o
d
f
o
r
tr
ai
n
in
g
an
d
test
in
g
o
f
v
eh
icle
ac
ce
ler
ated
au
to
m
atic
d
r
iv
in
g
alg
o
r
ith
m
.
T
h
e
s
y
s
tem
m
ain
ly
attem
p
ts
to
s
im
u
la
te
p
ed
estrian
h
ea
d
co
llis
io
n
test
r
e
s
u
lts
,
ef
f
ec
ti
v
en
ess
o
f
co
llis
io
n
war
n
in
g
m
o
d
el
an
d
co
llis
io
n
av
o
id
an
ce
f
o
r
in
tellig
en
t
co
n
n
ec
ted
v
eh
icles.
T
h
e
s
y
s
tem
in
[
1
6
]
attem
p
ts
f
o
r
ea
r
ly
d
e
tectio
n
o
f
p
ed
estrian
s
o
v
e
r
b
li
n
d
s
p
o
ts
th
at
ca
u
s
es
th
e
d
r
iv
er
to
b
lo
ck
th
e
v
iew.
T
h
e
s
y
s
tem
u
s
es
m
o
n
o
cu
lar
d
ep
th
esti
m
atio
n
alg
o
r
ith
m
th
at
allo
ws
a
s
m
aller
s
ea
r
ch
ar
ea
f
o
r
p
ed
estrian
d
ete
ctio
n
.
T
h
is
r
ed
u
ce
s
th
e
laten
cy
o
f
d
etec
tin
g
u
n
ex
p
ec
te
d
p
ed
e
s
tr
ian
s
.
I
t
h
as
b
ee
n
ex
p
er
im
en
ted
in
th
e
co
m
p
le
x
s
tr
ee
t
s
ce
n
es,
an
d
th
e
ac
cu
r
ac
y
s
ee
m
s
to
b
e
g
o
o
d
.
T
h
e
e
x
p
er
im
en
tal
r
esu
lts
r
ev
ea
led
th
at
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
m
eth
o
d
in
d
etec
tin
g
b
li
n
d
s
p
o
ts
an
d
s
u
d
d
en
a
p
p
ea
r
an
ce
o
f
p
ed
estrian
s
is
g
o
o
d
an
d
h
as
r
ed
u
ce
d
t
h
e
r
e
ac
tio
n
laten
cy
f
o
r
s
u
c
h
em
e
r
g
en
cy
s
itu
atio
n
s
b
y
ap
p
ly
in
g
th
e
p
r
e
-
d
etec
tio
n
m
eth
o
d
.
T
h
is
m
eth
o
d
co
m
p
a
r
ed
to
p
r
ev
io
u
s
m
eth
o
d
s
h
a
s
r
ed
u
ce
d
th
e
ac
cid
en
t
r
ate
ca
u
s
ed
b
y
s
u
d
d
en
p
ed
estrian
cr
o
s
s
in
g
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l time
o
b
ject
d
etec
tio
n
fo
r
a
d
va
n
ce
d
d
r
iver a
s
s
is
ta
n
ce
s
ystems
…
(
S
u
d
a
r
s
h
a
n
S
iva
k
u
m
a
r
)
4945
A
d
v
an
c
e
d
d
r
i
v
er
a
s
s
i
s
t
a
n
ce
s
y
s
t
e
m
s
en
h
an
c
e
v
eh
i
c
l
e
s
af
e
ty
b
y
u
t
i
l
i
z
i
n
g
s
en
s
o
r
s
a
n
d
in
te
l
l
i
g
e
n
t
ly
a
s
s
i
s
t
t
h
e
d
r
iv
e
r
s
i
n
n
a
v
ig
a
t
io
n
a
n
d
p
ar
k
in
g
t
a
s
k
s
.
T
h
e
s
e
s
y
s
t
e
m
s
i
n
co
r
p
o
r
a
t
e
c
a
m
e
r
a
-
b
a
s
e
d
s
en
s
o
r
s
t
o
i
n
c
r
e
a
s
e
d
r
iv
e
r
a
w
a
r
en
e
s
s
o
f
t
h
e
s
u
r
r
o
u
n
d
i
n
g
e
n
v
i
r
o
n
m
e
n
t.
C
e
n
t
r
a
l
t
o
A
D
A
S
f
u
n
c
t
i
o
n
a
l
it
y
a
r
e
s
y
s
t
em
s
-
on
-
a
-
c
h
ip
(
S
o
C
s
)
,
wh
i
c
h
in
t
eg
r
at
e
m
u
l
t
i
p
le
c
h
ip
s
t
o
c
o
n
n
e
c
t
s
e
n
s
o
r
s
w
i
t
h
a
c
tu
a
t
o
r
s
t
h
r
o
u
g
h
in
t
e
r
f
a
c
e
s
an
d
h
i
g
h
-
p
e
r
f
o
r
m
a
n
c
e
e
l
e
c
t
r
o
n
ic
c
o
n
tr
o
l
u
n
i
t
s
(
E
C
U
s
)
.
T
h
i
s
in
t
e
g
r
a
t
io
n
en
a
b
le
s
r
e
a
l
-
t
im
e
d
a
t
a
p
r
o
c
e
s
s
i
n
g
an
d
q
u
i
ck
r
e
s
p
o
n
s
e,
f
a
c
il
i
t
a
t
i
n
g
f
ea
t
u
r
e
s
s
u
ch
a
s
a
d
a
p
t
iv
e
c
r
u
i
s
e
c
o
n
tr
o
l
,
la
n
e
d
e
p
a
r
tu
r
e
w
ar
n
in
g
s
,
an
d
co
l
l
i
s
i
o
n
a
v
o
id
a
n
c
e
[
1
7
]
.
C
o
n
v
o
l
u
t
io
n
a
l
n
eu
r
a
l
n
e
t
wo
r
k
s
(
C
N
N
s
)
p
l
a
y
a
c
r
u
c
i
a
l
r
o
l
e
in
o
b
j
e
c
t
d
e
t
e
c
t
i
o
n
f
o
r
A
D
AS
s
p
e
c
i
f
i
c
a
l
ly
f
o
r
t
h
e
s
e
a
p
p
l
i
ca
t
i
o
n
s
,
w
i
th
d
a
t
a
s
e
t
s
s
u
ch
as
C
I
F
A
R
-
1
0
0
b
e
in
g
w
id
e
l
y
u
s
e
d
f
o
r
t
r
a
in
i
n
g
a
n
d
e
v
a
l
u
a
t
io
n
[
1
8
]
,
[
1
9
]
.
T
h
e
s
e
d
a
t
a
s
e
t
s
p
r
o
v
i
d
e
d
iv
e
r
s
e
i
m
a
g
e
c
a
te
g
o
r
i
e
s
,
e
n
a
b
l
i
n
g
r
o
b
u
s
t
m
o
d
e
l
g
e
n
er
a
l
i
z
a
t
io
n
.
T
h
e
wo
r
k
in
[
2
0
]
s
u
g
g
ests
th
at
Ad
v
an
ce
d
r
iv
er
ass
is
tan
t
s
y
s
tem
s
r
eq
u
ir
e
h
ig
h
co
m
p
u
tatio
n
al
p
o
wer
s
to
id
en
tify
an
d
class
if
y
th
e
o
b
jects
f
o
r
i
d
en
tify
in
g
th
e
a
u
g
m
en
ted
r
ea
lity
.
T
h
e
au
t
h
o
r
s
h
er
e
tr
y
to
cr
ea
te
lo
ca
lized
p
o
in
ts
ar
o
u
n
d
th
e
s
u
r
r
o
u
n
d
in
g
en
v
ir
o
n
m
en
t
s
u
ch
th
at
v
eh
icles
ca
p
tu
r
e
th
ese
o
b
jects
at
h
ig
h
er
s
p
ee
d
s
.
T
h
e
aim
is
t
o
cr
ea
te
a
th
r
ee
-
d
im
en
s
io
n
al
m
o
d
el
w
ith
ac
cu
r
ate
in
f
o
r
m
atio
n
ab
o
u
t
th
e
test
s
ite.
I
t
is
ex
p
ec
ted
th
at
u
s
in
g
AI
an
d
I
S
will
s
ig
n
if
ican
tly
im
p
r
o
v
e
p
e
r
f
o
r
m
a
n
ce
as
co
m
p
u
tatio
n
al
s
p
ee
d
an
d
ac
cu
r
ac
y
f
o
r
AR
ap
p
licatio
n
s
in
au
to
m
o
b
iles
.
I
n
th
e
wo
r
k
p
r
o
p
o
s
ed
i
n
[
2
1
]
,
th
e
au
th
o
r
s
tr
y
to
co
m
p
ar
e
au
to
m
ated
lan
e
k
ee
p
in
g
s
y
s
tem
s
(
AL
KS)
wit
h
th
e
ad
ap
tiv
e
cr
u
is
e
co
n
tr
o
l
(
AC
C
)
.
T
h
e
r
esu
lt
s
p
r
o
v
e
th
at
AL
KS
p
er
f
o
r
m
s
m
u
ch
b
etter
t
h
an
AC
C
.
T
h
e
r
a
p
id
a
d
v
an
ce
m
en
t
o
f
ar
t
if
icial
in
tellig
en
ce
(
AI
)
in
au
t
o
n
o
m
o
u
s
d
r
iv
in
g
h
as
le
d
to
s
ig
n
if
ican
t
ad
v
an
ce
s
,
en
a
b
lin
g
t
h
e
d
e
v
elo
p
m
en
t
o
f
r
o
b
u
s
t
d
r
iv
er
ass
is
tan
t
s
y
s
tem
s
[
2
2
]
.
Ho
wev
e
r
,
as
th
ese
s
y
s
tem
s
b
ec
o
m
e
m
o
r
e
p
r
e
v
alen
t,
ad
d
r
e
s
s
in
g
th
e
eth
ical
co
n
s
id
er
atio
n
s
s
u
r
r
o
u
n
d
i
n
g
th
eir
d
ep
lo
y
m
e
n
t
an
d
o
p
er
atio
n
is
cr
itical.
T
h
is
r
esear
ch
d
en
o
u
n
ce
s
th
e
m
u
ltifa
ce
ted
d
o
m
ain
o
f
eth
ics
in
AI
f
o
r
ADAS,
an
aly
zin
g
v
ar
io
u
s
u
s
e
ca
s
es
an
d
s
ce
n
ar
io
s
.
E
th
ical
c
o
n
ce
r
n
s
en
c
o
m
p
ass
to
p
ics
s
u
ch
as
s
af
ety
an
d
p
r
iv
ac
y
r
elate
d
to
d
ata
c
o
llectio
n
an
d
u
s
ag
e,
d
ec
is
io
n
-
m
a
k
in
g
d
ilem
m
as,
ac
co
u
n
tab
ilit
y
,
an
d
s
o
cieta
l
im
p
ac
t.
T
h
e
s
tu
d
y
f
o
cu
s
es
o
n
in
tr
icate
ch
allen
g
es
in
au
to
n
o
m
o
u
s
d
r
i
v
in
g
,
e
x
am
in
in
g
r
ea
l
-
wo
r
l
d
c
ases
to
s
h
ed
lig
h
t
o
n
c
o
m
p
lex
eth
ical
is
s
u
es.
T
h
e
p
ap
er
also
f
o
cu
s
es o
n
Data
s
ec
u
r
ity
wh
en
d
ata
is
tr
an
s
f
er
r
ed
f
r
o
m
o
n
e
ca
r
t
o
an
o
t
h
er
.
T
h
e
r
is
e
o
f
s
o
f
twar
e
-
d
ef
i
n
ed
v
eh
icles
(
SDVs
)
[
2
3
]
h
as
r
ap
id
l
y
en
h
an
ce
d
th
e
ad
v
a
n
ce
m
en
t
o
f
ADAS
,
au
to
n
o
m
o
u
s
v
e
h
icles
(
AVs),
an
d
b
atter
y
elec
tr
ic
v
eh
icle
(
B
E
V)
tech
n
o
lo
g
y
.
W
h
ile
AV
s
r
eq
u
ir
e
p
o
wer
to
co
m
p
u
te
d
ata
f
r
o
m
p
er
ce
p
tio
n
to
co
n
tr
o
ls
,
B
E
Vs
n
ee
d
ef
f
icien
cy
to
o
p
tim
ize
th
eir
elec
tr
ic
d
r
iv
in
g
r
an
g
e
an
d
s
tan
d
o
u
t
co
m
p
ar
e
d
to
tr
ad
itio
n
al
in
ter
n
al
co
m
b
u
s
tio
n
en
g
in
e
(
I
C
E
)
v
eh
icles.
AVs
p
o
s
s
es
s
ce
r
tain
lag
s
in
th
e
cu
r
r
en
t
wo
r
ld
,
b
u
t
SAE
L
ev
el
2
+
(
L
2
+
)
a
u
to
m
ated
v
eh
icl
es
ar
e
th
e
f
o
c
u
s
o
f
m
ajo
r
o
r
ig
in
al
eq
u
ip
m
en
t
m
an
u
f
ac
tu
r
er
s
(
OE
Ms)
.
T
h
e
co
m
m
o
n
f
o
r
m
o
f
an
SDV
to
d
ay
co
m
b
in
es
AV
an
d
B
E
V
tech
n
o
lo
g
y
o
n
th
e
s
am
e
p
latf
o
r
m
,
p
r
o
m
in
e
n
tly
av
ailab
le
in
m
o
s
t
OE
Ms
'
lin
eu
p
s
.
As
th
e
co
m
p
u
te
an
d
s
en
s
o
r
ar
c
h
itectu
r
es
f
o
r
L
2
+
au
to
m
ated
v
eh
icles
lean
to
w
ar
d
s
a
c
o
m
p
u
tatio
n
ally
e
x
p
e
n
s
iv
e
r
o
b
u
s
t
d
esig
n
,
it
m
ay
h
am
p
e
r
t
h
e
m
o
s
t
im
p
o
r
tan
t
p
u
r
c
h
asin
g
f
ac
to
r
o
f
a
B
E
V
-
th
e
elec
tr
ic
d
r
iv
in
g
r
an
g
e.
T
h
e
wo
r
k
d
escr
ib
e
d
in
[
2
4
]
tr
ies
to
ad
v
a
n
ce
th
e
YOL
O
d
etec
tio
n
ac
cu
r
ac
y
an
d
in
teg
r
atin
g
v
o
ice
-
b
ased
t
ec
h
n
iq
u
es
f
o
r
o
b
ject
d
etec
tio
n
.
T
h
e
p
ap
e
r
m
ain
l
y
f
o
cu
s
es
o
n
d
y
n
am
ic
co
n
v
o
lu
tio
n
s
an
d
atten
tio
n
m
ec
h
an
is
m
s
.
T
h
e
wo
r
k
in
[
2
5
]
d
escr
ib
es
th
e
ac
cu
r
ac
y
im
p
r
o
v
em
e
n
t
f
o
r
au
to
m
o
tiv
e
ap
p
licatio
n
s
f
o
r
d
if
f
e
r
en
t
YOL
O
ar
ch
itectu
r
es.
Her
e
th
e
p
ap
er
u
s
es
an
im
ag
e
p
r
e
-
p
r
o
ce
s
s
in
g
b
lo
c
k
a
n
d
cite
s
th
e
lim
itatio
n
s
o
f
p
r
e
v
io
u
s
l
y
u
s
ed
YOL
O
m
o
d
els.
T
h
e
r
esu
lt
in
th
is
p
ap
e
r
s
h
o
ws an
im
p
r
o
v
ed
o
b
ject
class
if
icatio
n
ac
cu
r
ac
y
.
I
n
th
e
p
r
o
p
o
s
ed
wo
r
k
in
t
h
is
p
ap
er
,
we
h
av
e
f
o
cu
s
ed
o
n
s
o
m
e
o
f
th
e
d
r
awb
ac
k
s
o
f
e
x
is
tin
g
o
b
ject
d
etec
tio
n
in
ADAS.
Ma
jo
r
c
o
n
tr
ib
u
tio
n
s
o
f
th
is
wo
r
k
i
n
clu
d
e
C
NN
an
d
R
NN
-
b
ased
o
b
ject
d
etec
tio
n
o
p
tim
ized
f
o
r
p
o
o
r
v
is
ib
ilit
y
co
n
d
itio
n
s
,
au
to
m
atic
b
r
ak
i
n
g
s
y
s
tem
in
teg
r
ated
with
A
B
S
E
C
U,
r
eal
-
tim
e
p
ed
estrian
an
d
d
r
iv
er
aler
ts
u
s
in
g
h
az
ar
d
lig
h
ts
an
d
h
o
n
k
in
g
,
i
m
p
r
o
v
ed
d
etec
tio
n
ac
cu
r
ac
y
(
7
8
.
3
%
f
o
r
u
n
clea
r
im
ag
es)
with
o
u
t im
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
an
d
ev
alu
atio
n
u
s
in
g
HI
L
s
im
u
latio
n
s
.
3.
P
RO
P
O
SE
D
M
E
T
H
O
DO
L
O
G
Y
Pro
p
o
s
ed
s
y
s
tem
in
Fig
u
r
e
1
in
teg
r
ates
a
h
y
b
r
id
m
o
d
el,
co
m
b
in
in
g
C
NNs
an
d
R
NNs
e
m
p
lo
y
ed
to
p
er
f
o
r
m
o
b
ject
class
if
icatio
n
an
d
tem
p
o
r
al
tr
ac
k
in
g
.
T
h
e
C
NN
ex
tr
ac
ts
s
p
atial
f
ea
tu
r
es
f
r
o
m
ea
ch
f
r
am
e,
wh
ile
th
e
R
NN
ca
p
tu
r
es
tem
p
o
r
al
d
ep
en
d
en
cies,
th
u
s
e
n
h
an
cin
g
d
etec
tio
n
p
er
f
o
r
m
an
ce
in
d
y
n
am
ic
d
r
iv
in
g
s
ce
n
es.
T
h
e
m
o
d
el
is
tr
ain
ed
an
d
test
ed
u
s
in
g
C
I
FAR
-
1
0
,
C
I
FAR
-
1
0
0
an
d
YOL
O
d
atas
ets,
ea
ch
co
n
tain
i
n
g
lo
w
-
r
eso
lu
tio
n
im
a
g
es
ac
r
o
s
s
1
0
an
d
1
0
0
ca
teg
o
r
ies,
r
esp
ec
tiv
ely
.
Desp
ite
th
e
d
ataset
lim
itatio
n
s
,
th
e
m
o
d
el
ac
h
iev
es
an
ac
cu
r
ac
y
o
f
7
8
.
3
%
with
o
u
t
p
r
ep
r
o
ce
s
s
in
g
f
o
r
C
I
FAR
-
1
0
5
0
%
f
o
r
C
I
FAR
-
1
0
0
d
ataset
an
d
9
3
%
f
o
r
YOL
O
v
8
d
ataset,
d
em
o
n
s
t
r
atin
g
r
o
b
u
s
tn
ess
to
r
aw
in
p
u
t
d
ata.
An
em
b
ed
d
e
d
s
y
s
tem
in
teg
r
a
tio
n
wh
er
e
th
e
tr
ain
e
d
m
o
d
el
is
d
ep
lo
y
ed
o
n
a
co
m
p
u
tin
g
p
latf
o
r
m
co
n
n
ec
ted
to
th
e
v
e
h
icle’
s
AB
S
E
C
U.
T
h
e
o
b
ject
d
etec
tio
n
o
u
tp
u
t
is
u
s
ed
to
tr
ig
g
er
b
r
a
k
i
n
g
co
m
m
a
n
d
s
b
ased
o
n
p
r
e
d
ef
in
ed
t
h
r
ea
t
ass
ess
m
en
t
cr
iter
ia.
T
h
is
s
tep
en
s
u
r
e
s
th
at
th
e
s
y
s
tem
tr
an
s
itio
n
s
f
r
o
m
p
e
r
ce
p
tio
n
t
o
ac
tio
n
,
a
k
ey
r
eq
u
ir
em
e
n
t
f
o
r
r
ea
l
-
wo
r
ld
ADAS
d
ep
lo
y
m
en
t.
T
o
ev
alu
ate
r
ea
l
-
tim
e
p
er
f
o
r
m
a
n
ce
an
d
r
eliab
ilit
y
,
th
e
en
tire
s
y
s
tem
is
test
ed
u
s
in
g
HI
L
s
im
u
latio
n
.
T
h
is
s
etu
p
en
ab
les
th
e
test
in
g
o
f
E
C
U
r
esp
o
n
s
es
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
4
2
-
4
9
5
3
4946
to
b
e
d
etec
ted
o
b
jects
u
n
d
e
r
co
n
tr
o
l
an
d
r
e
p
ea
tab
le
s
ce
n
ar
i
o
s
.
T
h
e
HI
L
en
v
ir
o
n
m
e
n
t
m
i
m
ics
ac
tu
al
v
e
h
icle
d
y
n
am
ics an
d
s
en
s
o
r
in
p
u
ts
,
allo
win
g
f
o
r
co
m
p
r
eh
en
s
iv
e
v
alid
atio
n
b
ef
o
r
e
in
-
v
eh
icle
d
ep
lo
y
m
en
t.
Fig
u
r
e
1
.
Ov
e
r
v
iew
o
f
th
e
p
r
o
p
o
s
ed
s
y
s
tem
ar
ch
itectu
r
e
T
o
im
p
r
o
v
e
ac
cu
r
ac
y
,
th
e
f
o
ll
o
win
g
m
o
d
if
icatio
n
s
wer
e
ap
p
lied
to
th
e
C
NN
ar
ch
itectu
r
e:
−
Den
s
e
l
ay
er
: I
n
cr
ea
s
ed
to
1
0
2
4
u
n
its
f
o
r
im
p
r
o
v
ed
f
ea
tu
r
e
e
x
tr
ac
tio
n
.
−
L
ea
r
n
in
g
r
ate:
Set to
0
.
0
1
with
an
ad
ap
tiv
e
d
ec
ay
f
u
n
ctio
n
.
−
Pad
d
in
g
&
Ker
n
el
c
o
n
s
tr
ain
ts
:
Op
tim
ized
f
o
r
b
etter
im
ag
e
f
e
atu
r
e
r
eten
tio
n
.
−
Use o
f
d
en
s
e
C
NN:
E
n
h
an
ce
d
class
if
icatio
n
p
er
f
o
r
m
an
ce
o
v
e
r
s
tan
d
ar
d
C
NN
m
o
d
els
−
Den
s
e
l
a
y
e
r
:
I
n
cr
ea
s
ed
t
o
1
0
2
4
u
n
i
ts
f
o
r
i
m
p
r
o
v
e
d
f
e
at
u
r
e
e
x
t
r
a
cti
o
n
.
−
L
ea
r
n
i
n
g
r
at
e:
Set
t
o
0
.
0
1
w
it
h
a
n
a
d
a
p
ti
v
e
d
e
ca
y
f
u
n
cti
o
n
.
−
Pad
d
i
n
g
&
Ke
r
n
el
c
o
n
s
t
r
ai
n
ts
:
Op
t
im
i
ze
d
f
o
r
b
e
tte
r
im
a
g
e
f
ea
tu
r
e
r
ete
n
t
io
n
.
−
Use
o
f
d
e
n
s
e
C
NN
:
E
n
h
an
ce
d
class
i
f
i
ca
t
io
n
p
e
r
f
o
r
m
an
ce
o
v
e
r
s
t
a
n
d
a
r
d
C
NN
m
o
d
els
A
2
D
co
n
v
o
lu
tio
n
al
lay
e
r
tr
an
s
f
o
r
m
s
an
in
p
u
t
f
ea
tu
r
e
m
ap
∈
×
×
\
{
}
^
{
\
\
}
∈
u
s
in
g
a
k
er
n
el
co
n
s
tr
ain
t K
in
o
r
d
er
t
o
p
r
o
d
u
ce
an
o
u
tp
u
t
f
e
atu
r
e
Yi,
=
=
0
∑
−
1
=
0
∑
−
1
=
0
∑
−
1
,
,
⋅
+
,
+
,
(
1
)
Fro
m
(
1
)
th
e
k
er
n
el
co
n
s
tr
ain
t
∥
K
∥
≤
λ
,
λ
=0
.
1
,
is
u
s
ed
to
r
eg
u
lar
ize
lear
n
in
g
,
b
alan
ce
th
e
u
n
b
alan
ce
d
weig
h
ts
an
d
p
r
e
v
en
t o
v
er
f
itti
n
g
o
f
th
e
m
o
d
el
wh
ile
g
o
in
g
lay
e
r
b
y
lay
er
.
T
h
e
o
u
tp
u
t d
im
en
s
io
n
o
f
a
f
ea
tu
r
e
m
ap
is
ca
lcu
lated
u
s
in
g
=
⌊
+
2
−
⌋
+
1
(
2
)
I
n
(
2
)
th
e
p
ad
d
in
g
ad
d
s
an
e
x
tr
a
0
o
r
1
b
ef
o
r
e
im
ag
e
p
ix
el
f
o
r
3
2
×
3
2
C
NN
n
etwo
r
k
i
n
o
r
d
e
r
to
p
r
ev
e
n
t
p
r
o
b
lem
s
at
th
e
e
d
g
es
o
f
t
h
e
C
NN
an
d
to
m
ain
tain
th
e
s
p
at
ial
s
id
e
at
th
e
o
u
tp
u
t.
A
d
em
o
o
f
v
alu
e
s
elec
tio
n
is
as
:
−
I
n
p
u
t size:
3
2
×
3
2
−
Ker
n
el:
3
−
Strid
e:
1
,
Fro
m
(
2
)
,
p
a
d
d
in
g
:
=
1
,
=
⌊
32
+
2
(
1
)
−
31
⌋
+
1
=
32
_
{
\
{
}
}
=
\
\
\
{
32
+
2
(
1
)
−
3
}
{
1
}
=
⌊
132
+
2
(
1
)
−
3
⌋
+
1
=
32
So
th
e
o
u
tp
u
t
r
etain
s
t
h
e
o
r
ig
in
al
s
ize.
L
ea
r
n
in
g
r
ate
d
ec
ay
is
in
v
er
s
e
o
f
tim
e
d
ec
ay
,
wh
ich
ad
ju
s
ts
th
e
lear
n
in
g
r
at
e
\
at
ea
ch
ep
o
ch
o
r
s
tep
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l time
o
b
ject
d
etec
tio
n
fo
r
a
d
va
n
ce
d
d
r
iver a
s
s
is
ta
n
ce
s
ystems
…
(
S
u
d
a
r
s
h
a
n
S
iva
k
u
m
a
r
)
4947
=
01
+
\
_
=
\
{
\
_
0
}
{
1
+
\
ℎ
}
=
1
+
0
(
3
)
Fro
m
th
e
(
3
)
th
e
d
ec
ay
f
u
n
ctio
n
in
th
e
ar
ch
itectu
r
e
is
u
s
ed
to
a
d
ap
t
with
th
e
n
o
is
y
r
aw
in
p
u
ts
,
p
a
r
ticu
lar
ly
with
o
u
t
th
e
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
c
k
s
.
T
h
is
m
ak
es
v
er
y
h
i
g
h
u
p
d
ates
in
th
e
tr
ain
in
g
d
ata
to
m
in
im
ize
lo
ca
l
m
in
im
u
m
,
u
s
ef
u
l f
o
r
co
n
v
er
g
e
n
ce
an
d
s
tab
ilit
y
4.
DATAS
E
T
DE
SCRI
P
T
I
O
N,
P
RE
P
RO
CE
SS
I
NG
AND
T
RAINI
NG
T
h
e
C
I
FAR
-
1
0
an
d
C
I
FA
R
-
1
0
0
d
atasets
ar
e
wid
ely
u
s
ed
b
e
n
ch
m
ar
k
s
in
co
m
p
u
ter
v
is
io
n
task
s
.
B
o
th
d
atasets
co
n
s
is
t
o
f
co
lo
r
ed
n
at
u
r
al
im
ag
es,
ea
ch
with
a
r
eso
l
u
tio
n
o
f
3
2
×
3
2
p
ix
els.
C
I
FAR
-
1
0
i
n
clu
d
es 6
0
,
0
0
0
im
ag
es
ca
teg
o
r
ized
in
to
1
0
cl
ass
es
s
u
ch
as
ca
r
s
,
tr
u
ck
s
,
air
p
lan
es,
an
d
an
im
als,
wh
ile
C
I
FAR
-
1
0
0
co
m
p
r
is
es
th
e
s
am
e
n
u
m
b
e
r
o
f
im
ag
es b
u
t a
cr
o
s
s
1
0
0
f
in
e
-
g
r
ai
n
ed
ca
teg
o
r
ies.
Alth
o
u
g
h
th
ese
d
atasets
d
o
n
o
t
h
a
v
e
tr
af
f
ic
s
ce
n
ar
io
s
,
th
ey
d
e
m
o
n
s
tr
ate
e
f
f
ec
tiv
e
te
s
tb
ed
s
f
o
r
ev
alu
atin
g
m
o
d
el
ar
ch
itectu
r
e
,
tr
ain
in
g
ef
f
icien
cy
,
an
d
o
b
je
ct
r
ec
o
g
n
itio
n
ca
p
ab
ilit
ies
d
u
r
in
g
p
o
o
r
v
is
ib
ilit
y
an
d
ad
v
e
r
s
e
wea
th
er
co
n
d
itio
n
s
.
C
lass
e
s
s
u
ch
as
au
to
m
o
b
ile
an
d
tr
u
ck
i
n
C
I
FAR
-
1
0
p
r
o
v
id
e
lim
ited
y
et
r
elev
an
t
r
ele
v
an
ce
f
o
r
v
eh
icu
l
ar
o
b
ject
r
ec
o
g
n
itio
n
with
lo
w
er
r
eso
lu
tio
n
s
t
o
ev
al
u
ate
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
m
o
d
el.
All
i
n
p
u
t
i
m
ag
es
ar
e
n
o
r
m
alize
d
to
en
s
u
r
e
u
n
if
o
r
m
it
y
an
d
im
p
r
o
v
e
co
n
v
er
g
e
n
ce
d
u
r
in
g
tr
ain
in
g
.
T
h
e
p
ix
el
v
alu
es
ar
e
s
ca
led
to
th
e
r
an
g
e
[
0
,
1
]
,
an
d
d
ata
au
g
m
en
tatio
n
tech
n
iq
u
es
s
u
ch
as
h
o
r
izo
n
tal
f
lip
p
i
n
g
,
cr
o
p
p
in
g
,
an
d
r
o
tatio
n
ar
e
a
p
p
lied
to
im
p
r
o
v
e
g
e
n
er
aliza
tio
n
an
d
r
ed
u
ce
o
v
er
f
itti
n
g
.
T
h
e
YOL
O
v
8
d
ataset
co
n
s
is
ts
o
f
im
ag
es lik
e
s
to
p
,
y
ield
,
s
p
e
ed
lim
it,
p
ed
estrian
cr
o
s
s
in
g
,
n
o
en
tr
y
&
U
-
tu
r
n
(
le
f
t/
r
ig
h
t)
.
T
h
is
is
f
i
r
s
t
p
r
ep
r
o
ce
s
s
ed
u
s
in
g
Ultr
a
ly
tics
YOL
Ov
8
f
r
am
ewo
r
k
,
wh
er
e
th
e
o
b
jects
r
ep
r
esen
ted
in
tx
t
f
o
r
m
at
ar
e
co
n
v
er
ted
t
o
b
in
ar
y
v
ar
iab
les
o
f
0
’
s
an
d
1
’
s
.
On
ce
th
e
im
ag
e
p
ix
els
ar
e
co
n
v
er
ted
to
th
is
f
o
r
m
at,
th
e
m
o
d
el
is
ac
co
r
d
in
g
ly
tr
ain
e
d
to
o
b
tain
th
e
b
o
u
n
d
i
n
g
b
o
x
in
th
e
(
,
&
f
o
r
m
at)
to
id
en
tify
th
e
ex
ac
t p
o
s
itio
n
s
in
th
e
r
ea
l
-
wo
r
l
d
v
alid
atio
n
.
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
Fo
llo
win
g
p
er
f
o
r
m
an
ce
m
etr
ic
s
ar
e
u
s
ed
f
o
r
an
aly
s
is
:
Dete
ctio
n
s
u
cc
ess
r
ate:
7
8
.
3
%
(
C
I
FAR
-
1
0
)
,
5
0
%
(
C
I
FAR
-
1
0
0
)
f
o
r
u
n
clea
r
im
ag
e
s
am
p
les
an
d
9
3
%
f
o
r
YOL
O
v
8
with
o
u
t u
s
in
g
im
a
g
e
p
r
ep
r
o
ce
s
s
in
g
b
o
ck
.
Sto
p
p
in
g
d
is
tan
ce
: 5
m
eter
s
b
e
f
o
r
e
co
llis
io
n
.
L
aten
cy
r
e
d
u
ctio
n
:
Fas
ter
h
az
ar
d
r
esp
o
n
s
e
o
v
er
p
r
e
v
io
u
s
m
o
d
e.
I
n
au
t
o
m
o
tiv
e
em
b
ed
d
ed
s
y
s
tem
s
th
e
ac
cu
r
ac
y
o
f
ca
p
tu
r
in
g
an
d
p
r
o
ce
s
s
in
g
u
n
clea
r
im
a
g
es
u
s
in
g
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
is
d
ef
in
ed
as
a
p
r
o
p
o
r
tio
n
o
f
tr
u
e
p
o
s
itiv
e
d
etec
tio
n
s
am
o
n
g
all
th
e
p
o
s
itiv
e
d
etec
tio
n
s
.
I
n
ca
s
e
o
f
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
,
all
p
o
s
itiv
e
d
etec
tio
n
s
f
o
r
d
etec
tin
g
an
u
n
clea
r
im
ag
e
ar
e
6
0
,
0
0
0
im
ag
e
s
am
p
les
o
u
t
o
f
wh
ich
4
6
,
9
8
0
im
ag
e
s
am
p
les
ar
e
d
etec
ted
wh
ic
h
is
th
e
tr
u
e
p
o
s
itiv
e
d
etec
tio
n
f
o
r
C
I
FAR
-
1
0
d
ataset.
Fo
r
C
I
FAR
-
1
0
0
d
ataset
to
tal
p
o
s
itiv
e
d
etec
tio
n
s
ar
e
6
0
,
0
0
0
im
ag
e
s
am
p
les
an
d
tr
u
e
p
o
s
itiv
e
d
etec
tio
n
s
ar
e
3
0
,
0
0
0
im
a
g
e
s
am
p
les.
T
a
b
le
1
p
r
o
v
id
es a
c
o
m
p
ar
is
o
n
o
f
e
x
is
tin
g
an
d
p
r
o
p
o
s
ed
d
r
iv
er
ass
is
tan
ce
s
y
s
tem
s
.
Fro
m
T
ab
le
1
it
is
v
er
y
ev
id
en
t
th
at
t
h
e
p
r
o
p
o
s
ed
s
y
s
tem
p
er
f
o
r
m
s
well
b
y
u
s
in
g
th
e
d
en
s
e
co
n
v
o
l
u
tio
n
n
e
u
r
al
n
etwo
r
k
s
with
o
u
t
an
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
.
On
e
m
ajo
r
im
p
r
o
v
em
en
t
h
er
e
is
th
e
s
y
s
tem
also
aler
ts
th
e
p
ed
estri
an
wh
er
ein
th
e
p
r
ev
io
u
s
s
y
s
tem
o
n
ly
aler
ts
th
e
d
r
iv
er
.
W
e
a
ls
o
s
ee
h
er
e
th
at
th
e
d
etec
tio
n
ac
cu
r
ac
y
is
in
cr
ea
s
ed
with
o
u
t
an
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
in
th
e
ad
v
er
s
e
wea
th
er
co
n
d
itio
n
s
an
d
th
e
s
y
s
tem
s
to
p
s
au
to
m
atica
lly
if
th
e
d
r
i
v
er
is
n
o
t r
esp
o
n
d
i
n
g
f
o
r
f
ew
s
ec
o
n
d
s
.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
ex
is
tin
g
an
d
p
r
o
p
o
s
ed
d
r
iv
er
ass
is
tan
ce
s
y
s
tem
F
e
a
t
u
r
e
Ex
i
s
t
i
n
g
a
d
v
a
n
c
e
d
d
r
i
v
e
r
a
ss
i
st
a
n
c
e
s
y
st
e
ms
P
r
o
p
o
se
d
a
d
v
a
n
c
e
d
d
r
i
v
e
r
a
ss
i
st
a
n
c
e
s
y
st
e
m
O
b
j
e
c
t
t
y
p
e
s
Li
mi
t
e
d
t
o
p
e
d
e
st
r
i
a
n
s
I
n
c
l
u
d
e
s
c
a
t
s,
a
u
t
o
mo
b
i
l
e
s,
t
r
u
c
k
s
C
N
N
a
r
c
h
i
t
e
c
t
u
r
e
S
t
a
n
d
a
r
d
D
e
n
se
C
N
N
I
mag
e
-
p
r
e
p
r
o
c
e
ss
i
n
g
b
l
o
c
k
s
R
e
q
u
i
r
e
d
N
o
t
r
e
q
u
i
r
e
d
A
u
t
o
ma
t
i
c
b
r
a
k
i
n
g
N
o
t
p
r
e
s
e
n
t
P
r
e
sen
t
v
i
a
A
B
S
EC
U
i
n
t
e
g
r
a
t
i
o
n
I
mag
e
p
r
e
p
r
o
c
e
ssi
n
g
R
e
q
u
i
r
e
d
N
o
t
r
e
q
u
i
r
e
d
P
e
d
e
s
t
r
i
a
n
a
l
e
r
t
s
D
r
i
v
e
r
o
n
l
y
D
r
i
v
e
r
+
P
e
d
e
st
r
i
a
n
D
e
t
e
c
t
i
o
n
a
c
c
u
r
a
c
y
7
5
.
8
%
(
c
l
e
a
r
i
m
a
g
e
s)
3
0
%
f
o
r
C
I
F
A
R
-
1
0
0
(
u
n
c
l
e
a
r
s
a
m
p
l
e
s
)
7
8
.
3
%
(
u
n
c
l
e
a
r
i
ma
g
e
s)
5
0
%
(
f
o
r
C
I
F
A
R
-
1
0
0
u
n
c
l
e
a
r
s
a
m
p
l
e
s
)
W
e
ac
h
iev
ed
a
n
ac
c
u
r
ac
y
o
f
7
8
.
3
%
f
o
r
b
lu
r
r
e
d
im
a
g
es,
wh
e
r
e
th
e
ac
cu
r
ac
y
was
ab
o
u
t
7
5
.
3
%
[
6
]
f
o
r
clea
r
im
ag
e
s
am
p
le
in
r
elate
d
wo
r
k
.
I
t
is
wo
r
th
m
en
ti
o
n
in
g
th
at
th
e
ac
cu
r
ac
y
o
f
7
8
.
3
%
i
s
ac
h
iev
ed
with
o
u
t
u
s
in
g
an
y
p
r
e
-
p
r
o
ce
s
s
in
g
b
lo
ck
s
.
I
f
p
r
e
-
p
r
o
ce
s
s
in
g
b
l
o
ck
s
ar
e
u
s
ed
th
e
ac
cu
r
ac
y
will
b
e
m
u
ch
g
r
ea
ter
f
o
r
C
I
FAR
1
0
d
ataset.
Fo
r
C
I
FA
R
1
0
0
d
ataset
th
e
ac
cu
r
ac
y
is
3
4
.
3
8
%
f
o
r
5
ep
o
c
h
s
.
Usi
n
g
d
en
s
e
n
eu
r
al
n
etwo
r
k
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
4
2
-
4
9
5
3
4948
we
wer
e
ab
le
to
ac
h
iev
e
5
0
%
ac
cu
r
ac
y
an
d
f
o
r
YOL
O
d
ataset
(
r
ea
l
tim
e
au
to
m
o
tiv
e
d
atas
et)
we
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
3
%
f
o
r
8
ep
o
c
h
s
[
2
3
]
–
[
2
5
]
.
T
ab
le
2
d
e
m
o
n
s
t
r
ates
th
ese
r
esu
lts
.
Fig
u
r
e
s
2
,
3
,
4
a
n
d
T
ab
le
3
clea
r
ly
in
d
icate
th
e
r
esu
lts
o
f
o
u
r
wo
r
k
o
u
t
p
er
f
o
r
m
s
co
m
p
a
r
ed
to
p
r
e
v
io
u
s
ly
o
b
tain
ed
r
esu
lts
[
1
4
]
,
[
1
5
]
,
[
2
3
]
–
[
2
5
]
.
T
ab
le
s
2
an
d
3
g
iv
e
th
e
ac
cu
r
ac
y
e
f
f
ec
tiv
en
ess
o
f
p
r
e
v
io
u
s
r
esu
lts
v
/s
th
e
r
esu
lts
o
b
tain
ed
b
y
c
h
an
g
in
g
th
e
s
tr
u
ctu
r
e.
T
ab
le
2
.
Acc
u
r
ac
y
af
ter
c
h
an
g
in
g
th
e
C
NN
ar
ch
itectu
r
e
A
c
c
u
r
a
c
y
f
o
r
C
I
F
A
R
-
1
0
d
a
t
a
se
t
u
si
n
g
C
N
N
[
1
3
]
–
[
1
5
]
,
[
2
2
]
A
c
c
u
r
a
c
y
f
o
r
C
I
F
A
R
-
1
0
d
a
t
a
se
t
u
si
n
g
C
N
N
A
c
c
u
r
a
c
y
o
f
C
I
F
A
R
1
0
0
d
a
t
a
se
t
u
si
n
g
D
e
n
se
C
N
N
[
1
3
]
–
[
1
5
]
,
[
2
2
]
A
c
c
u
r
a
c
y
o
f
C
I
F
A
R
1
0
0
u
s
i
n
g
d
e
n
se
C
N
N
68
-
7
5
.
8
%
7
8
.
3
%
3
4
.
3
8
%
5
0
%
Fig
u
r
e
2
.
Acc
u
r
ac
y
u
s
in
g
C
NN
n
etwo
r
k
f
o
r
C
I
FAR
-
10
Fig
u
r
e
3
.
Acc
u
r
ac
y
u
s
in
g
C
NN
n
etwo
r
k
f
o
r
C
I
FAR
1
0
0
with
o
u
t a
n
y
ch
an
g
es
Fig
u
r
e
4
.
Acc
u
r
ac
y
u
s
in
g
d
en
s
e
C
NN
af
ter
m
o
d
if
icatio
n
o
f
a
r
ch
itectu
r
e
T
ab
le
3
.
C
o
m
p
a
r
is
o
n
o
f
r
esu
lts
(
with
an
d
with
o
u
t im
a
g
e
p
r
e
p
r
o
ce
s
s
in
g
b
lo
c
k
)
D
a
t
a
s
e
t
M
o
d
e
l
U
se
o
f
i
ma
g
e
p
r
e
p
r
o
c
e
ssi
n
g
b
l
o
c
k
A
c
c
u
r
a
c
y
F
1
s
c
o
r
e
C
I
F
A
R
-
10
C
N
N
Y
e
s
68
-
7
5
.
8
%
0
.
6
8
-
0
.
7
5
8
C
I
F
A
R
-
10
C
N
N
No
7
8
.
3
%
0
.
7
8
3
C
I
F
A
R
-
100
D
e
n
se
R
C
N
N
Y
e
s
3
4
.
3
8
%
0
.
3
3
C
I
F
A
R
1
0
0
D
e
n
se
R
C
N
N
NO
5
0
%
0
.
4
9
Y
O
LO
V
8
C
N
N
Y
e
s
85
-
9
0
%
0
.
8
5
-
0
.
9
Y
O
LO
V
8
C
N
N
No
9
3
%
0
.
9
3
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l time
o
b
ject
d
etec
tio
n
fo
r
a
d
va
n
ce
d
d
r
iver a
s
s
is
ta
n
ce
s
ystems
…
(
S
u
d
a
r
s
h
a
n
S
iva
k
u
m
a
r
)
4949
T
ab
le
3
p
r
o
v
id
es
th
e
ac
cu
r
a
cy
an
d
F1
s
co
r
e
b
ased
o
n
th
e
p
ar
am
eter
s
o
f
s
elec
tio
n
o
f
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
s
an
d
th
e
alg
o
r
ith
m
u
s
ed
.
T
h
e
r
esu
lts
p
r
o
v
e
th
at
th
e
p
r
o
p
o
s
ed
s
y
s
tem
h
as
ac
h
iev
ed
g
r
ea
ter
ac
cu
r
ac
y
with
o
u
t
th
e
u
s
ag
e
o
f
an
im
a
g
e
p
r
e
p
r
o
ce
s
s
in
g
b
lo
ck
f
o
r
b
o
t
h
C
I
FAR
1
0
C
I
FAR
1
0
0
an
d
YOL
O
d
ataset
b
y
m
o
d
if
y
in
g
th
e
s
tr
u
ctu
r
e.
Fro
m
th
e
o
b
tain
ed
r
esu
lts
it c
an
b
e
o
b
s
er
v
ed
th
at
th
e
s
y
s
tem
is
ex
p
ec
ted
to
s
to
p
b
ef
o
r
e
5
m
o
f
d
etec
tin
g
an
o
b
ject
wh
en
th
e
u
n
clea
r
im
ag
es
ar
e
u
s
ed
.
T
h
e
s
y
s
tem
r
esp
o
n
s
ib
le
f
o
r
th
is
ac
tio
n
is
ca
lled
h
ar
d
war
e
in
lo
o
p
s
y
s
tem
(
HI
L
)
th
at
is
in
te
g
r
ated
with
th
e
em
b
ed
d
e
d
s
y
s
tem
as
s
h
o
wn
in
Fig
u
r
e
5
.
A
HI
L
s
y
s
tem
in
th
e
au
to
m
o
tiv
e
is
an
em
b
ed
d
e
d
test
b
en
ch
wh
er
e
all
th
e
m
o
d
u
les
ar
e
in
teg
r
ated
to
g
eth
er
in
a
clo
s
ed
l
o
o
p
a
n
d
th
e
r
esp
o
n
s
e
o
f
th
e
s
y
s
tem
is
ev
alu
ated
b
y
v
alid
atin
g
th
e
co
r
r
ec
tn
ess
o
f
an
alg
o
r
ith
m
f
o
r
a
s
p
ec
if
ic
d
ataset.
T
h
e
HI
L
s
y
s
tem
p
r
o
p
o
s
ed
in
Fig
u
r
e
5
co
n
s
is
ts
o
f
4
AR
S
6
2
0
R
ADAR
s
en
s
o
r
m
o
d
u
le,
4
SVC
3
0
0
ca
m
er
a
,
an
d
an
AB
S
E
C
U
th
at
is
co
n
n
ec
ted
to
a
m
aster
co
n
tr
o
l
E
C
U.
T
h
e
d
ataset
an
d
alg
o
r
ith
m
a
r
e
f
lash
ed
in
to
th
e
s
e
ca
m
er
as
with
th
e
h
elp
o
f
a
tech
n
iq
u
e
ca
lled
PP
AR
f
lash
in
g
u
s
in
g
T
r
ac
e
3
2
d
ev
ice
th
at
is
in
teg
r
ated
with
th
e
HI
L
s
y
s
tem
.
All
th
ese
s
e
n
s
o
r
m
o
d
u
les
an
d
E
C
U
is
th
e
in
p
u
t
to
th
e
b
r
ak
e
to
r
q
u
e
m
o
d
u
le
o
f
t
h
e
em
b
ed
d
ed
s
y
s
tem
.
T
h
e
b
r
ak
e
to
r
q
u
e
is
ad
ju
s
ted
b
ased
o
n
th
e
ca
m
er
a’
s
in
p
u
t
o
n
d
etec
tin
g
a
p
e
d
estrian
o
r
an
o
b
ject
with
th
e
h
elp
o
f
R
ADAR
s
en
s
o
r
s
.
I
t
ca
n
also
b
e
in
f
er
r
e
d
f
r
o
m
th
e
r
esu
lts
th
at
th
e
s
en
s
itiv
ity
r
esp
o
n
s
e
o
f
th
e
s
y
s
tem
is
g
o
o
d
as
o
b
s
er
v
ed
f
r
o
m
t
h
e
r
esu
lts
s
h
o
wn
in
T
ab
le
3
,
Fig
u
r
e
s
2
,
4
,
6
,
an
d
7
[
6
]
,
[
1
4
]
,
[
1
5
]
,
[
2
3
]
–
[
2
5
]
.
Fig
u
r
e
5
.
Pro
p
o
s
ed
e
m
b
ed
d
e
d
s
y
s
tem
d
esig
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
5
,
Octo
b
e
r
20
25
:
4
9
4
2
-
4
9
5
3
4950
Fig
u
r
e
6
.
Dete
ctio
n
a
cc
u
r
ac
y
o
f
YOL
O
V8
,
(
c
lass
1
5
-
s
p
ee
d
l
im
its
,
class
1
1
-
B
ik
es,
p
ed
estrian
an
d
B
icy
cles)
Fig
u
r
e
7
.
R
esu
lts
d
escr
ib
in
g
o
r
ig
in
al
an
d
im
p
r
o
v
ed
ac
c
u
r
ac
y
Fro
m
th
e
r
esu
lts
an
d
ar
c
h
itectu
r
al
ch
an
g
es
d
o
n
e
in
th
e
n
e
u
r
al
n
etwo
r
k
s
in
ac
co
r
d
a
n
ce
wit
h
a
m
ajo
r
ar
ch
itectu
r
e
c
h
an
g
e
in
AB
S
E
C
U,
it
ca
n
b
e
o
b
s
er
v
ed
th
at
we
h
av
e
r
ec
eiv
e
d
a
n
ac
c
u
r
ac
y
o
f
3
.
3
%
m
o
r
e
th
a
n
th
e
p
r
ev
io
u
s
wo
r
k
f
o
r
C
I
FAR
-
1
0
d
ataset
u
s
in
g
co
n
v
o
lu
tio
n
n
eu
r
al
n
etwo
r
k
s
[
6
]
,
[
1
4
]
,
[
1
5
]
,
1
6
%
m
o
r
e
f
o
r
C
I
FAR
-
100
[
1
0
]
–
[
1
5
]
f
o
r
u
n
c
lear
d
atasets
an
d
2
%
[
2
3
]
–
[
2
5
]
m
o
r
e
f
o
r
YOL
Ov
8
im
ag
es
d
u
r
in
g
b
a
d
wea
th
er
co
n
d
itio
n
s
with
o
u
t
u
s
in
g
a
n
i
m
ag
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
c
k
.
Fr
o
m
th
e
p
ast
wo
r
k
it
ca
n
b
e
o
b
s
er
v
ed
th
at
f
o
r
clea
r
im
ag
e
s
am
p
les
u
s
in
g
an
im
a
g
e
p
r
ep
r
o
ce
s
s
in
g
b
lo
ck
th
e
ac
cu
r
ac
y
was
ab
o
u
t
6
8
%
an
d
u
s
in
g
d
en
s
e
C
NN
it
was
ar
o
u
n
d
3
0
%
in
all
th
e
p
r
ev
io
u
s
r
esear
ch
wo
r
k
s
cited
in
th
e
p
ap
er
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
is
ex
p
ec
ted
to
s
to
p
with
in
5
m
eter
s
.
5
.
1
.
Adv
a
nced
e
m
bedd
ed
s
y
s
t
em
HI
L
v
alid
atio
n
was
co
n
d
u
cte
d
with
4
AR
S
6
2
0
R
ADAR
Sen
s
o
r
s
,
4
SVC
3
0
0
C
am
er
as
,
AB
S
E
C
U
co
n
n
ec
ted
to
a
Ma
s
ter
C
o
n
tr
o
l
E
C
U
an
d
PP
A
R
Flas
h
in
g
f
o
r
d
ataset
in
teg
r
atio
n
.
R
esu
lts
co
n
f
ir
m
e
n
h
an
ce
d
b
r
ak
in
g
r
esp
o
n
s
e
tim
e
a
n
d
i
m
p
r
o
v
e
d
d
etec
tio
n
r
eliab
ilit
y
in
ch
allen
g
i
n
g
c
o
n
d
itio
n
s
.
T
h
e
4
-
c
h
an
n
el
b
ased
An
til
-
lo
ck
b
r
ak
in
g
s
y
s
tem
m
o
d
u
le
(
u
s
u
ally
B
OSC
H
A
B
S
/ESP
co
n
tr
o
l
m
o
d
u
le)
in
Fig
u
r
e
5
c
o
n
s
is
ts
o
f
wh
ee
l
s
p
ee
d
s
en
s
o
r
s
m
o
u
n
ted
o
n
4
co
r
n
er
s
o
f
th
e
wh
ee
l,
Hy
d
r
au
lic
v
alv
es,
B
r
ak
e
Pu
m
p
an
d
th
e
E
C
U
alo
n
g
with
ca
m
er
a
s
en
s
o
r
s
an
d
ad
ap
tiv
e
cr
u
is
e
co
n
tr
o
l
s
y
s
tem
.
T
h
e
v
iew
p
o
r
t
E
C
U
an
d
th
e
AB
S
E
C
U
ar
e
co
n
n
ec
ted
to
a
m
aster
E
C
U
u
s
in
g
C
AN
p
r
o
to
co
l th
at
wo
r
k
s
b
ased
o
n
th
e
m
u
ltima
s
ter
-
m
u
ltis
lav
e
p
r
o
to
co
l
.
Als
o
,
th
e
ADAS
s
y
s
tem
co
n
s
is
t
s
o
f
ca
m
e
r
a
s
en
s
o
r
s
th
at
ar
e
m
o
u
n
ted
o
n
th
e
f
o
u
r
co
r
n
er
s
o
f
th
e
o
b
ject.
T
h
e
m
ain
wo
r
k
o
f
th
e
C
NN
an
d
d
en
s
e
n
eu
r
al
n
etwo
r
k
alg
o
r
ith
m
in
teg
r
ated
in
th
e
c
am
er
a
s
en
s
o
r
is
to
class
if
y
th
e
ty
p
e
o
f
o
b
ject
an
d
d
etec
t th
e
d
is
tan
ce
o
f
th
e
b
lu
r
r
ed
s
am
p
le
esp
ec
ially
d
u
r
i
n
g
th
e
p
o
o
r
v
is
ib
ilit
y
o
r
u
n
clea
r
wea
th
e
r
co
n
d
itio
n
s
.
W
h
en
th
e
ca
m
e
r
a
s
en
s
o
r
d
ete
cts
a
p
er
s
o
n
,
(
f
o
r
e
x
am
p
le
wi
th
in
1
0
m
eter
s
)
,
th
e
ca
m
er
a
s
en
s
o
r
s
s
en
d
a
p
u
ls
e
s
ig
n
al
to
th
e
s
p
ee
d
s
en
s
o
r
s
to
d
ec
r
ea
s
e
th
e
s
p
ee
d
as
p
er
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
.
T
h
e
s
p
ee
d
is
d
ec
r
ea
s
ed
with
th
e
h
elp
o
f
Ad
ap
tiv
e
c
r
u
is
e
c
o
n
tr
o
l.
T
h
e
d
r
iv
er
s
tar
ts
g
ettin
g
an
aler
t
on
th
e
d
ash
b
o
ar
d
p
ar
allelly
to
ap
p
ly
b
r
ak
es.
T
h
e
m
aster
co
n
tr
o
l
E
C
U
p
r
esen
t
in
th
e
s
y
s
tem
g
r
ad
u
ally
r
ed
u
ce
s
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
R
ea
l time
o
b
ject
d
etec
tio
n
fo
r
a
d
va
n
ce
d
d
r
iver a
s
s
is
ta
n
ce
s
ystems
…
(
S
u
d
a
r
s
h
a
n
S
iva
k
u
m
a
r
)
4951
s
p
ee
d
with
th
e
h
elp
o
f
AB
S
s
y
s
tem
s
wh
er
e
th
e
h
y
d
r
au
lic
v
alv
es
au
to
m
atica
lly
co
n
tr
o
l
th
e
r
elea
s
e
an
d
ac
tio
n
o
f
th
e
b
r
ak
e
f
lu
id
s
ac
co
r
d
in
g
t
o
a
u
n
if
o
r
m
tim
e
in
ter
v
al
b
etwe
en
th
e
d
is
tan
ce
o
f
th
e
o
b
jec
t.
T
h
is
is
d
o
n
e
with
th
e
h
elp
o
f
h
y
d
r
a
u
lic
v
alv
es.
W
h
en
th
e
d
r
i
v
er
is
ig
n
o
r
an
t
til
l
th
e
8
th
m
eter
,
th
e
s
y
s
tem
au
t
o
m
atica
lly
g
o
es
to
a
h
alt
s
tate
wh
er
e
th
e
AB
S
m
o
d
u
le
p
r
esen
t
r
ec
eiv
es
a
s
ig
n
al
to
s
to
p
th
e
ca
r
i
n
2
m
eter
s
f
r
o
m
th
e
m
aster
E
C
U,
th
er
eb
y
a
u
to
m
atica
lly
tu
r
n
in
g
o
n
th
e
h
az
ar
d
lig
h
ts
an
d
h
o
r
n
b
u
zz
er
s
.
Fro
m
th
e
T
ab
le
2
it
is
clea
r
th
at
th
e
ac
c
u
r
ac
y
o
f
d
etec
tin
g
o
b
jects
h
as
in
cr
ea
s
ed
b
y
3
.
3
%
f
o
r
C
I
FAR
-
1
0
d
ataset
f
o
r
b
lu
r
r
ed
/
u
n
clea
r
s
am
p
les
wh
en
it
is
f
o
u
n
d
to
b
e
a
m
ax
im
u
m
o
f
6
8
%
f
o
r
clea
r
s
am
p
les
its
elf
[
6
]
,
[
1
4
]
,
[
1
5
]
.
T
h
is
is
e
v
id
en
t
f
r
o
m
th
e
r
esu
lts
o
f
Fig
u
r
e
7
.
Fr
o
m
Fig
u
r
e
7
it
is
o
b
s
er
v
e
d
th
at
th
e
ac
c
u
r
ac
y
o
f
C
I
FAR
1
0
0
is
3
4
%
wh
en
C
NN
is
ap
p
lied
,
b
u
t
in
o
u
r
ca
s
e,
we
h
a
v
e
ac
h
iev
ed
5
0
%
wh
ic
h
is
1
6
%
m
o
r
e
b
y
ap
p
ly
in
g
d
en
s
e
C
NN
an
d
f
r
o
m
Fig
u
r
e
6
it
i
s
v
e
r
y
ev
i
d
en
t
th
at
th
e
d
etec
tio
n
ac
cu
r
ac
y
9
3
%
f
o
r
YOL
O
v
8
d
ataset,
wh
ile
it wa
s
9
0
% in
th
e
r
ef
er
en
ce
s
cited
p
r
ev
io
u
s
ly
in
T
ab
le
3
[
2
3
]
–
[
2
5
]
.
T
h
is
was
p
o
s
s
ib
le
b
ec
au
s
e
o
f
th
e
p
r
o
p
o
s
ed
em
b
ed
d
ed
s
y
s
te
m
wh
er
ein
th
e
h
y
d
r
a
u
lic
ac
tio
n
o
f
b
r
a
k
e
f
lu
id
s
an
d
s
ig
n
al
f
r
o
m
th
e
s
p
e
ed
s
en
s
o
r
s
ar
e
au
to
m
atica
lly
c
o
n
tr
o
lled
with
r
esp
ec
t
to
th
e
s
p
ec
if
ic
tim
e
in
ter
v
al
b
ased
o
n
th
e
d
is
tan
ce
o
f
th
e
o
b
ject
u
s
in
g
b
r
ak
e
v
alv
es.
B
u
t
i
n
th
e
p
r
ev
io
u
s
wo
r
k
s
,
th
e
ac
tio
n
o
f
s
p
ee
d
s
en
s
o
r
s
an
d
h
y
d
r
au
lic
m
ec
h
an
is
m
in
a
n
ti
-
lo
ck
b
r
a
k
in
g
E
C
U
is
tr
ig
g
er
ed
o
n
l
y
wh
en
th
e
d
r
iv
er
p
r
e
s
s
es
th
e
b
r
ak
e
p
ad
s
u
s
in
g
f
u
zz
y
lo
g
ic
a
n
d
o
t
h
er
al
g
o
r
ith
m
s
[
1
]
–
[
3
]
.
6.
CO
NCLU
SI
O
N
AND
F
UR
U
RE
SCO
P
E
T
h
is
p
ap
er
p
r
esen
ts
a
r
o
b
u
s
t A
d
v
an
ce
d
r
iv
e
r
ass
is
tan
ce
s
y
s
te
m
th
at
en
h
an
ce
s
o
b
ject
d
etec
tio
n
in
r
ea
l
-
wo
r
ld
d
r
iv
in
g
s
ce
n
ar
io
s
lev
e
r
a
g
in
g
C
NN
an
d
d
e
n
s
e
n
eu
r
al
n
etwo
r
k
s
,
v
alid
ated
th
r
o
u
g
h
HI
L
s
im
u
latio
n
s
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
ad
d
r
ess
es
cr
itical
ADA
S
b
y
en
s
u
r
in
g
r
o
b
u
s
t
p
ed
estrian
d
etec
tio
n
,
au
to
m
atic
b
r
ak
in
g
,
an
d
r
ea
l
-
tim
e
h
az
ar
d
ale
r
ts
.
Un
lik
e
co
n
v
e
n
tio
n
al
m
et
h
o
d
s
,
w
h
ich
r
ely
o
n
d
r
iv
er
in
ter
v
e
n
tio
n
,
th
is
s
y
s
tem
au
to
n
o
m
o
u
s
ly
s
to
p
s
th
e
v
eh
i
cle,
en
h
an
ci
n
g
r
o
ad
s
af
ety
an
d
ad
v
a
n
ce
d
d
r
iv
er
ass
is
tan
ce
r
eliab
ilit
y
with
th
e
cu
s
to
m
er
s
.
B
y
in
teg
r
atin
g
d
ee
p
lear
n
in
g
tech
n
iq
u
es
with
AB
S
E
C
U
au
to
m
atio
n
,
th
e
s
y
s
tem
s
ig
n
if
ican
tly
im
p
r
o
v
es
s
af
ety
with
o
u
t
r
e
q
u
i
r
in
g
ad
d
itio
n
al
im
ag
e
p
r
e
p
r
o
c
ess
in
g
b
lo
ck
in
v
er
y
b
a
d
/ad
v
e
r
s
e
en
v
ir
o
n
m
en
tal
ch
allen
g
es.
I
t
h
as
also
im
b
ib
e
d
s
o
m
e
o
f
th
e
s
en
s
o
r
f
u
s
io
n
t
ec
h
n
iq
u
es
as
well.
B
y
elim
i
n
atin
g
t
h
e
n
ee
d
f
o
r
im
ag
e
p
r
ep
r
o
ce
s
s
in
g
,
th
e
s
y
s
tem
ac
h
iev
es
r
eliab
le
p
e
r
f
o
r
m
an
ce
u
n
d
er
a
d
v
er
s
e
v
is
ib
ilit
y
co
n
d
itio
n
s
,
with
d
etec
tio
n
ac
cu
r
ac
y
o
f
7
8
.
3
%
(
C
I
FAR
-
1
0
)
,
5
0
%
(
C
I
FAR
-
1
0
0
)
,
a
n
d
9
3
%
(
YOL
Ov
8
)
.
T
h
e
in
teg
r
atio
n
o
f
au
to
m
atic
b
r
ak
in
g
th
r
o
u
g
h
th
e
AB
S
E
C
U
f
u
r
th
er
en
h
an
ce
s
v
eh
icu
lar
s
af
ety
b
y
r
ed
u
cin
g
d
e
p
en
d
en
ce
o
n
d
r
iv
e
r
in
ter
v
en
tio
n
.
T
h
is
d
em
o
n
s
tr
at
es
th
e
s
y
s
tem
's
r
ea
l
-
wo
r
ld
ap
p
licab
ilit
y
an
d
r
o
b
u
s
tn
ess
.
Ho
wev
er
,
th
e
u
s
e
o
f
C
I
FAR
d
ata
s
ets p
o
s
es lim
ita
ti
o
n
o
f
lack
o
f
tr
af
f
ic
-
s
p
ec
if
ic
c
o
n
tex
t.
Fu
tu
r
e
wo
r
k
wo
u
ld
f
o
cu
s
o
n
r
ea
l
-
wo
r
ld
v
eh
icle
d
ep
lo
y
m
en
t
f
o
r
f
u
r
th
er
test
in
g
.
M
u
lti
-
m
o
d
al
s
en
s
o
r
f
u
s
io
n
(
L
I
DAR,
R
ADAR,
T
h
er
m
al
C
am
er
as)
ca
n
e
n
h
an
ce
d
etec
tio
n
ac
cu
r
ac
y
an
d
i
m
p
lem
en
tatio
n
o
f
T
r
an
s
f
o
r
m
e
r
-
b
ased
v
is
io
n
m
o
d
els
ca
n
r
esu
lt
i
n
b
etter
im
ag
e
class
if
icatio
n
.
Als
o
,
th
e
al
g
o
r
ith
m
ca
n
b
e
test
ed
f
o
r
tr
ain
in
g
an
d
ev
al
u
atio
n
o
f
th
e
m
o
d
el
o
n
r
ea
l
-
wo
r
ld
ADAS
d
ataset
s
s
u
ch
as
KI
T
T
I
,
B
DD1
0
0
K
o
r
n
u
Scen
es,
alo
n
g
with
i
n
teg
r
at
in
g
th
e
d
etec
tio
n
p
ip
elin
e
i
n
to
an
em
b
ed
d
ed
h
ar
d
war
e
p
la
tf
o
r
m
f
o
r
r
ea
l
-
tim
e
test
in
g
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
is
to
d
ec
lar
e
th
at
th
er
e
is
n
o
f
u
n
d
in
g
g
iv
en
to
t
h
is
wo
r
k
b
y
an
y
o
f
th
e
f
o
r
u
m
s
an
d
th
is
s
ec
tio
n
is
n
o
t a
p
p
licab
le.
A
UT
H
O
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Su
d
ar
s
h
an
Siv
ak
u
m
ar
✓
✓
✓
✓
✓
✓
✓
✓
✓
Sh
ik
h
a
T
r
ip
ath
i
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.