I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
. 14, No. 5, O
c
to
be
r
2025
, pp.
4279
~
4289
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
5
.pp
4279
-
4289
4279
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
A
u
t
om
at
e
d
vi
al
d
e
f
e
c
t
i
n
sp
e
c
t
i
on
u
si
n
g
G
ab
or
w
ave
l
e
t
s
an
d
k
-
m
e
an
s
c
l
u
st
e
r
i
n
g
V
is
h
w
an
at
h
a C
. R
.
1
,
V
.
A
s
h
a
2
, C
h
an
n
ab
as
ava
1
, S
r
e
e
k
an
t
h
R
al
la
p
al
li
1
1
D
e
pa
r
t
m
e
nt
of
M
a
s
t
e
r
of
C
om
put
e
r
A
ppl
i
c
a
t
i
ons
,
N
i
t
t
e
M
e
e
na
ks
hi
I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy,
N
i
t
t
e
(
D
e
e
m
e
d t
o be
U
ni
ve
r
s
i
t
y)
,
B
e
nga
l
ur
u, I
ndi
a
2
D
e
pa
r
t
m
e
nt
of
M
a
s
t
e
r
of
C
om
put
e
r
A
ppl
i
c
a
t
i
ons
, N
e
w
H
or
i
z
on C
ol
l
e
g
e
of
E
ngi
ne
e
r
i
ng, B
e
nga
l
ur
u, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
O
c
t
10, 2024
R
e
vi
s
e
d
J
ul
1, 2025
A
c
c
e
pt
e
d
J
ul
13, 2025
This
study
proposes
a
machine
vision
-
based
defect
inspection
system
for
pharmaceutical
vials,
aiming
to
ensure
the
quality
and
safety
of
me
dicinal
fluids.
The
system
employs
a
series
of
image
processing
tech
niques,
including
denoising,
feature
extractio
n
using
the
Gabor
wavelet
tran
sform,
segmentation,
clustering
with
the
K
-
means
algorithm,
and
precise
defect
identifica
tion
using
the
Canny
edge
operator.
Experimental
results
demonstrate high performance
, with recall, precision, accur
acy, and F1
-
score
exceeding
98%.
Additi
onally,
the
proposed
method
achieves
area
un
der
the
curve
-
receiver
-
operating
characteristic
curve
(
AUC
-
ROC
)
and
AUC
-
precision
-
recall
(
PR
)
values
of
approximately
98%.
The
system'
s
a
verage
computat
ional t
ime is
355 mi
croseconds,
indicat
ing it
s potent
ial for real
-
time
defect
detection.
Overall,
this
approach
offers
an
effective
solution
for
identifying
various
cosmetic
defects
such
as
scratche
s,
bruises,
cracks,
and
black
spots,
in
pharmaceutical
vials
without
the
need
for
vial
clas
sif
ication
training.
K
e
y
w
o
r
d
s
:
G
a
bor
w
a
ve
le
ts
K
-
m
e
a
ns
c
lu
s
te
r
in
g
M
a
c
hi
ne
vi
s
io
n
S
e
gm
e
nt
a
ti
on
V
ia
l
de
f
e
c
t
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
V
.
A
s
ha
D
e
pa
r
tm
e
nt
of
M
a
s
te
r
of
C
om
put
e
r
A
ppl
ic
a
ti
ons
, N
e
w
H
or
iz
on C
ol
le
ge
of
E
ngi
ne
e
r
in
g
B
e
nga
lu
r
u
-
560103, Ka
r
na
ta
ka
, I
ndi
a
E
m
a
il
:
a
s
ha
.gur
uda
th
@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
S
a
f
e
,
hi
gh
-
qua
li
ty
vi
a
ls
a
r
e
vi
ta
l
f
or
m
a
in
ta
in
in
g
s
te
r
il
e
c
ondi
ti
ons
a
nd
pr
e
s
e
r
vi
ng
m
e
di
c
in
e
in
te
gr
it
y
in
pha
r
m
a
c
e
ut
ic
a
ls
.
T
he
y
s
hi
e
ld
f
lu
id
s
f
r
om
c
ont
a
m
in
a
ti
on,
l
e
a
ks
,
a
nd
de
gr
a
da
ti
on.
A
dva
n
c
e
d
in
s
p
e
c
ti
on
m
e
th
ods
li
ke
m
a
c
hi
ne
vi
s
io
n,
opt
ic
a
l
in
s
pe
c
ti
on,
a
nd
a
ut
om
a
te
d
de
f
e
c
t
de
te
c
ti
on
e
ns
ur
e
c
om
pl
ia
nc
e
w
it
h
qua
li
ty
s
ta
nda
r
ds
li
ke
good
m
a
nuf
a
c
tu
r
in
g
pr
a
c
ti
c
e
s
(
G
M
P
)
a
nd
good
la
bor
a
to
r
y
pr
a
c
ti
c
e
s
(
G
L
P
)
.
B
y
le
ve
r
a
gi
ng
te
c
hnol
ogy,
th
e
s
e
m
e
th
ods
d
e
te
c
t
s
ubt
le
de
f
e
c
ts
,
e
na
bl
in
g
c
or
r
e
c
ti
ve
a
c
ti
on
s
a
nd
pr
e
ve
nt
in
g
de
f
e
c
ti
ve
vi
a
ls
f
r
om
e
nt
e
r
in
g
th
e
m
a
r
ke
t.
E
vol
vi
ng
qua
li
ty
a
s
s
ur
a
nc
e
(
QA
)
pr
ot
oc
ol
s
a
nd
te
c
hnol
ogy
in
te
gr
a
ti
on,
pa
r
ti
c
ul
a
r
ly
m
a
c
hi
ne
le
a
r
ni
ng
(
ML
)
w
it
h
c
om
put
e
r
vi
s
io
n
(
CV
)
,
e
nha
nc
e
pha
r
m
a
c
e
ut
ic
a
l
c
ont
a
in
e
r
r
e
li
a
bi
li
ty
a
nd
s
a
f
e
ty
by
id
e
nt
if
yi
ng
is
s
ue
s
li
ke
c
r
a
c
ks
,
bl
a
c
k
s
pot
s
, s
c
r
a
tc
he
s
,
a
nd
bubble
s
.
F
ig
ur
e
1
s
how
s
a
t
ypi
c
a
l
vi
a
l
us
e
d i
n
pha
r
m
a
c
e
ut
ic
a
l
pa
c
ki
ng.
F
ig
ur
e
1. C
om
m
onl
y us
e
d
pha
r
m
a
c
e
ut
ic
a
l
vi
a
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 5, O
c
to
be
r
2025
:
4279
-
4289
4280
V
a
r
io
u
s
m
e
th
o
dol
ogi
e
s
ha
ve
be
e
n
a
ppl
ie
d
to
d
e
t
e
c
t
f
l
a
w
s
i
n
g
la
s
s
v
ia
ls
a
nd bot
tl
e
s
. T
h
e
l
ow
-
a
n
gl
e
a
nd
la
r
g
e
di
v
e
r
ge
nc
e
a
n
gl
e
(
L
A
L
D
A
)
vi
s
i
on
s
ys
te
m
[
1]
e
nh
a
n
c
e
s
ph
a
r
m
a
c
e
ut
i
c
a
l
b
ot
t
le
c
a
p
i
n
s
p
e
c
ti
on
by
im
pr
ov
in
g
de
f
e
c
t
c
o
nt
r
a
s
t
a
nd
m
i
ti
g
a
t
in
g
o
ve
r
e
x
po
s
ur
e
,
a
ddr
e
s
s
in
g
o
c
c
lu
s
i
on
s
a
nd
in
c
on
s
i
s
t
e
nt
il
lu
m
in
a
ti
on
.
u
s
i
ng
H
S
V
-
ba
s
e
d
m
u
lt
i
-
c
h
a
n
ne
l
s
e
g
m
e
nt
a
ti
on
(
M
C
S
)
,
it
a
c
hi
e
v
e
s
ov
e
r
9
5
%
a
c
c
ur
a
c
y.
X
-
r
a
y
-
b
a
s
e
d
in
s
pe
c
ti
on
s
y
s
t
e
m
[
2]
de
t
e
c
t
s
n
on
-
m
e
t
a
ll
ic
c
o
nt
a
m
i
na
nt
s
b
ut
r
e
li
e
s
on
c
o
s
tl
y
de
f
e
c
ti
ve
s
a
m
pl
e
s
.
T
hi
s
s
tu
dy
a
c
hi
e
v
e
s
97
.4%
a
c
c
u
r
a
c
y
f
or
gl
a
s
s
f
r
a
gm
e
nt
s
,
im
pr
o
vi
n
g
s
e
ns
it
i
vi
t
y
to
1.
0
a
nd
F
1
-
s
c
or
e
t
o
0.
980
.
H
e
u
r
i
s
ti
c
s
e
gm
e
nt
a
t
io
n
f
or
vi
a
l
d
e
f
e
c
t
de
t
e
c
ti
on
[
3]
s
h
ow
e
d
h
ig
h
a
c
c
ur
a
c
y
b
ut
r
e
qui
r
e
d
va
li
d
a
t
io
n
o
n
la
r
g
e
r
d
a
t
a
s
e
ts
.
F
a
r
ha
ng
i
e
t
a
l
.
[
4]
pr
op
os
e
d
a
th
r
e
s
hol
di
n
g
-
b
a
s
e
d
s
ys
te
m
f
or
de
te
c
t
in
g
c
a
p
de
f
e
c
t
s
,
s
e
a
l
in
te
gr
i
ty
,
a
nd
li
qui
d
le
ve
l
v
a
r
ia
t
io
ns
i
n
e
d
ib
l
e
oi
l
a
nd
be
v
e
r
a
g
e
bo
tt
l
e
s
,
a
c
h
ie
vi
n
g
95
.6%
a
c
c
ur
a
c
y
u
s
i
ng
a
c
om
pl
e
m
e
n
ta
r
y
m
e
ta
l
oxi
de
s
e
m
i
c
o
ndu
c
t
or
(
C
M
O
S
)
c
a
m
e
r
a
.
H
o
w
e
v
e
r
, w
or
n
l
a
b
e
l
s
a
n
d w
e
a
k s
e
a
l
s
p
o
s
e
c
h
a
l
le
ng
e
s
. L
iu
e
t
a
l
.
[
5]
a
c
hi
e
v
e
d m
or
e
t
ha
n
9
8%
a
c
c
u
r
a
c
y
in
d
e
te
c
t
in
g
vi
a
l
d
e
f
e
c
t
s
,
bu
t
f
a
c
e
d
c
ha
ll
e
ng
e
s
in
a
d
a
p
ti
n
g
to
d
if
f
e
r
e
n
t
li
g
ht
i
ng
c
on
di
t
io
n
s
.
T
he
bl
a
c
k
to
p
-
h
a
t
tr
a
n
s
f
or
m
a
lg
or
it
h
m
[
6]
c
o
m
bi
ne
d
w
it
h
hybr
id
im
a
g
e
pr
o
c
e
s
s
i
ng
a
d
dr
e
s
s
e
d
lo
c
a
li
z
a
ti
on
i
s
s
ue
s
b
ut
ha
d
hi
g
h
c
om
put
a
t
io
n
a
l
c
os
ts
,
i
m
p
a
c
ti
n
g
r
e
a
l
-
ti
m
e
d
e
pl
o
ym
e
nt
.
O
th
e
r
a
pp
r
oa
c
h
e
s
,
s
u
c
h
a
s
C
a
nn
y
e
dg
e
d
e
t
e
c
ti
on
[
7]
a
nd
c
lu
s
t
e
r
i
ng
m
e
th
od
s
[
8]
,
p
e
r
f
or
m
e
d
w
e
ll
bu
t
s
tr
ug
gl
e
d
w
i
th
s
pe
c
i
f
ic
d
e
f
e
c
t
s
,
w
hi
l
e
a
dv
a
n
c
e
d
m
ul
t
iv
i
e
w
s
y
s
t
e
m
s
[
9]
e
n
c
o
un
te
r
e
d
s
e
gm
e
n
ta
ti
o
n
c
ha
l
le
ng
e
s
du
e
to
un
e
v
e
n
l
ig
h
ti
n
g
.
P
la
s
t
ic
bot
tl
e
de
f
e
c
t
d
e
t
e
c
ti
o
n
[
10
]
pr
e
s
e
n
te
d
c
h
a
ll
e
n
ge
s
w
it
h
d
a
t
a
a
nd
c
o
m
p
ut
a
ti
o
na
l
n
e
e
d
s
,
w
hi
le
z
o
n
a
l
a
nd
ti
m
e
-
s
h
a
r
in
g
c
om
put
a
ti
on
a
l
im
a
gi
ng
(
Z
T
S
C
I
)
te
c
hn
iq
u
e
s
[
11]
l
a
i
d
a
f
ou
nd
a
ti
on
f
or
gr
a
y
im
a
g
e
f
l
a
w
d
e
t
e
c
ti
o
n.
C
o
nv
ol
u
ti
o
na
l
ne
ur
a
l
ne
two
r
k
s
(
C
N
N
s
)
a
nd
de
e
p
l
e
a
r
ni
ng
m
o
de
l
s
[
12]
,
[
13]
i
m
pr
ov
e
d
d
e
t
e
c
ti
o
n
c
a
p
a
b
il
i
ti
e
s
on
l
a
r
ge
d
a
t
a
s
e
t
s
,
w
hi
l
e
f
i
e
ld
-
pr
o
gr
a
m
m
a
bl
e
ga
t
e
a
r
r
a
y
(
F
P
G
A
)
-
ba
s
e
d
s
y
s
t
e
m
s
[
14]
a
c
c
e
l
e
r
a
t
e
d
bot
tl
e
c
a
p i
n
s
p
e
c
ti
o
n,
th
oug
h
s
t
il
l
f
a
c
e
d
s
e
r
v
e
r
s
tr
a
in
.
X
u
e
t
al
.
[
15]
pr
opos
e
d
a
de
f
e
c
t
de
te
c
ti
on
s
ys
te
m
f
or
f
il
le
d
vi
a
ls
th
a
t
in
te
gr
a
te
s
tr
a
di
ti
ona
l
im
a
ge
pr
oc
e
s
s
in
g
a
nd
de
e
p
le
a
r
ni
ng.
D
e
f
e
c
t
de
te
c
ti
on
of
s
ur
f
a
c
e
a
nd
c
ont
e
nt
s
in
vi
a
ls
(
D
D
S
C
N
e
t
)
,
bui
lt
on
Y
O
L
O
v8
w
it
h
qua
dr
a
f
us
io
n
a
nd
a
tt
e
nt
io
n
(
Q
U
F
U
A
tt
)
f
or
f
e
a
tu
r
e
f
us
io
n,
A
C
m
ix
f
or
de
f
e
c
t
f
oc
us
,
a
nd
li
ne
a
r
de
f
or
m
a
bl
e
c
onvolut
io
n
f
or
w
e
a
k
f
e
a
tu
r
e
e
xt
r
a
c
ti
on,
a
c
hi
e
ve
s
76.7%
m
A
P
on
V
ia
lG
1_D
E
T
,
65.9%
o
n
V
ia
lG
2_D
E
T
,
a
nd
86.9%
on
V
ia
lG
3_D
E
T
w
it
h
9.3
G
F
L
O
P
S
,
out
pe
r
f
or
m
in
g
Y
O
L
O
v11
by
3.5%
.
Y
O
L
O
-
ba
s
e
d
m
ode
ls
[
16]
–
[
22]
de
m
ons
tr
a
te
d
e
f
f
ic
ie
nt
de
f
e
c
t
de
te
c
ti
on
a
c
r
os
s
di
f
f
e
r
e
nt
s
ur
f
a
c
e
s
,
but
s
om
e
s
tr
uggl
e
d
w
it
h
s
m
a
ll
ta
r
ge
ts
or
hi
gh
c
om
put
a
ti
ona
l
c
os
ts
.
L
a
s
tl
y,
m
e
th
ods
us
in
g
r
obot
ic
te
c
hnol
ogi
e
s
a
nd
A
I
[
23]
,
[
24
]
e
nha
nc
e
d
in
s
pe
c
ti
on
a
c
c
ur
a
c
y,
th
ough
c
ha
ll
e
nge
s
w
i
th
s
e
ns
or
c
a
li
br
a
ti
on
a
nd
s
c
a
la
bi
li
ty
pe
r
s
is
te
d.
T
he
pr
opos
e
d
s
tu
dy
a
ddr
e
s
s
e
s
th
e
s
e
ga
ps
by
a
im
in
g
f
or
c
om
pr
e
he
ns
iv
e
de
f
e
c
t
de
te
c
ti
on
a
c
r
o
s
s
th
e
e
nt
ir
e
vi
a
l
s
ur
f
a
c
e
,
e
nha
n
c
in
g
r
e
li
a
bi
li
ty
w
it
hout
r
e
ly
in
g
on
c
om
pl
e
x
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
.
T
h
e
c
ur
a
te
d
da
ta
s
e
t
ta
il
or
e
d f
or
vi
a
l
de
f
e
c
t
in
s
pe
c
ti
on ove
r
c
om
e
s
l
im
it
a
ti
ons
s
e
e
n i
n
e
xi
s
ti
ng me
th
ods
.
T
he
m
ot
iv
a
ti
on
f
or
us
in
g
G
a
bor
w
a
ve
le
t
s
is
th
e
ir
a
bi
li
ty
to
d
e
te
c
t
vi
a
l
de
f
e
c
ts
vi
a
m
a
c
hi
ne
vi
s
io
n
w
it
hout
hum
a
n
in
te
r
ve
nt
io
n
[
25]
.
I
n
th
is
s
tu
dy,
G
a
bor
w
a
ve
le
ts
de
noi
s
e
a
nd
e
xt
r
a
c
t
vi
a
l
im
a
ge
f
e
a
tu
r
e
s
,
w
hi
c
h a
r
e
t
he
n s
e
gm
e
nt
e
d by the
K
-
m
e
a
ns
c
lu
s
te
r
in
g a
lg
or
it
hm
ba
s
e
d on f
e
a
tu
r
e
s
im
il
a
r
it
ie
s
. T
he
C
a
nny
e
dge
ope
r
a
to
r
hi
ghl
ig
ht
s
de
f
e
c
ts
a
c
r
os
s
th
e
vi
a
l'
s
s
ur
f
a
c
e
,
n
e
c
k,
to
p, a
nd
bot
to
m
w
it
h
hi
gh
pr
e
c
is
io
n.
P
yt
hon,
a
lo
ng
w
it
h
th
e
s
c
ik
it
im
a
ge
li
br
a
r
y,
is
us
e
d
f
or
im
pl
e
m
e
nt
a
ti
on.
K
e
y
f
e
a
tu
r
e
s
of
th
e
m
e
th
od
in
c
lu
de
it
s
a
ppl
ic
a
bi
li
ty
to
a
ny
de
f
e
c
t
in
a
ny
vi
a
l
r
e
gi
on,
m
a
ki
ng
it
hi
ghl
y
ve
r
s
a
ti
le
.
I
t
doe
s
not
r
e
qui
r
e
a
tr
a
in
in
g
s
ta
ge
or
de
f
e
c
t
-
f
r
e
e
s
a
m
pl
e
s
,
th
e
r
e
by
e
li
m
in
a
ti
ng
th
e
ne
e
d
f
or
th
r
e
s
hol
di
ng.
T
he
de
f
e
c
t
de
te
c
ti
on
pr
oc
e
s
s
is
f
ul
ly
a
ut
om
a
te
d
th
r
ough
K
-
m
e
a
ns
c
lu
s
te
r
in
g,
a
ll
ow
in
g
f
or
th
e
s
pe
c
if
ic
d
e
te
c
ti
on
of
va
r
io
us
de
f
e
c
ts
s
uc
h
a
s
s
c
r
a
tc
h
e
s
,
bubble
s
,
a
nd c
r
a
c
ks
, r
a
th
e
r
t
ha
n m
e
r
e
ly
c
a
t
e
gor
iz
in
g vi
a
ls
.
T
he
pa
p
e
r
is
or
ga
ni
z
e
d
in
th
e
f
ol
lo
w
in
g
w
a
y:
s
e
c
ti
on
2
br
ie
f
ly
e
xpl
a
in
s
th
e
pr
opo
s
e
d
m
ode
l
f
or
de
f
e
c
t
de
te
c
ti
on
us
in
g
G
a
bor
w
a
ve
le
t
s
a
nd
K
-
m
e
a
ns
c
lu
s
t
e
r
in
g.
S
e
c
ti
on
3
pr
e
s
e
nt
s
e
xpe
r
im
e
nt
on
va
r
io
us
r
e
a
l
de
f
e
c
ti
ve
vi
a
l
im
a
ge
s
.
T
he
c
onc
lu
s
io
ns
a
r
e
di
s
c
us
s
e
d i
n
s
e
c
ti
on
4.
2.
M
E
T
H
O
D
M
a
c
hi
ne
vi
s
io
n
is
us
e
d
in
pl
a
c
e
of
hum
a
n
vi
s
io
n
to
de
te
c
t
de
f
e
c
ts
us
in
g
c
om
put
e
r
a
lg
or
it
hm
s
a
nd
pr
oc
e
s
s
or
s
.
T
he
pr
opos
e
d
m
a
c
hi
ne
vi
s
io
n
s
y
s
te
m
c
om
pr
is
e
s
f
our
s
te
ps
.
F
ir
s
t,
a
n
im
a
ge
is
a
c
qui
r
e
d
u
s
in
g
a
hi
gh
-
r
e
s
ol
ut
io
n
in
dus
tr
ia
l
c
a
m
e
r
a
.
I
n
th
e
s
e
c
ond
s
te
p,
th
e
im
a
ge
unde
r
goe
s
pr
e
pr
oc
e
s
s
in
g,
w
he
r
e
it
is
de
-
noi
s
e
d, a
nd f
e
a
tu
r
e
s
s
u
c
h a
s
e
ne
r
gy, c
ont
r
a
s
t,
a
nd va
r
ia
nc
e
i
n m
ul
ti
pl
e
or
ie
nt
a
ti
ons
a
nd s
c
a
le
s
a
r
e
e
xt
r
a
c
te
d
us
in
g
th
e
G
a
bor
w
a
ve
le
t
tr
a
ns
f
or
m
.
T
he
th
ir
d
s
te
p
in
vol
ve
s
c
lu
s
te
r
in
g
a
nd
s
e
gm
e
nt
a
ti
on
of
th
e
e
xt
r
a
c
te
d
f
e
a
tu
r
e
s
to
hi
ghl
ig
ht
de
f
e
c
ts
.
F
in
a
ll
y,
a
de
c
is
io
n
is
m
a
de
ba
s
e
d
on
th
e
s
e
gm
e
nt
e
d
out
put
.
F
ig
ur
e
2
s
how
s
th
e
bl
oc
k di
a
gr
a
m
of
t
he
s
e
s
te
p
s
.
2.1. Gab
or
w
ave
le
t
s
T
he
G
a
bor
w
a
ve
le
t
is
a
m
a
th
e
m
a
ti
c
a
l
f
unc
ti
on
us
e
d
to
a
na
ly
z
e
im
a
ge
s
in
bot
h
s
pa
ti
a
l
a
nd
f
r
e
que
nc
y
dom
a
in
s
,
f
il
te
r
in
g
im
a
ge
s
vi
a
r
e
a
l
pa
r
ts
to
e
xt
r
a
c
t
pa
tt
e
r
ns
[
26]
.
I
n
vi
a
l
de
f
e
c
t
de
te
c
ti
on,
G
a
bor
w
a
ve
le
ts
a
r
e
a
ppl
ie
d
to
vi
a
l
im
a
ge
s
f
or
f
e
a
tu
r
e
e
xt
r
a
c
ti
on.
T
he
f
il
te
r
s
c
a
ns
pi
xe
l
va
lu
e
s
in
a
ll
di
r
e
c
ti
ons
,
id
e
nt
if
yi
ng
e
dge
da
ta
poi
nt
s
by
de
te
c
ti
ng
m
a
xi
m
um
va
lu
e
s
in
gr
a
di
e
nt
in
te
ns
it
y
m
a
tr
ix
,
e
na
bl
in
g
pa
tt
e
r
n
id
e
nt
if
ic
a
ti
on
a
nd
de
f
e
c
t
de
te
c
ti
on.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
ut
om
at
e
d v
ia
l
de
fe
c
t
in
s
pe
c
ti
on us
in
g
G
abo
r
w
av
e
l
e
ts
and
k
-
m
e
ans
c
lu
s
te
r
in
g
(
V
is
hw
anat
ha C
. R
.)
4281
F
iv
e
f
a
c
to
r
s
c
a
n
a
f
f
e
c
t
f
il
te
r
in
g
w
he
n
u
s
in
g
G
a
bor
w
a
ve
le
ts
.
T
o
c
om
pr
e
he
nd
th
is
,
c
ons
id
e
r
a
s
in
e
w
a
ve
ove
r
la
id
on
a
two
-
di
m
e
ns
io
na
l
G
a
us
s
ia
n
b
e
ll
c
ur
ve
.
S
in
c
e
it
'
s
a
two
-
di
m
e
ns
io
n
a
l
be
ll
c
ur
ve
,
th
e
s
in
e
w
a
ve
i
s
di
r
e
c
ti
ona
l
a
nd c
a
n pos
s
e
s
s
va
r
io
us
or
ie
nt
a
ti
ons
, t
hus
i
n
f
lu
e
nc
in
g t
he
f
il
te
r
w
it
h t
he
f
ol
lo
w
in
g
f
a
c
to
r
s
:
i)
D
ir
e
c
ti
on
(
θ
)
:
it
i
ndi
c
a
te
s
t
he
di
r
e
c
ti
on of
t
he
s
in
e
w
a
ve
ii)
O
f
f
s
e
t
(
φ)
:
in
di
c
a
te
s
t
he
pha
s
e
of
f
s
e
t
of
t
he
s
in
e
w
a
ve
iii)
S
ta
nda
r
d de
vi
a
ti
on
(
σ
)
:
th
e
s
m
a
ll
e
r
t
he
va
lu
e
s
of
t
hi
s
a
tt
r
ib
ut
e
, t
he
c
lo
s
e
r
t
he
va
lu
e
s
a
r
e
t
o t
he
c
e
nt
e
r
iv
)
E
ll
ip
ti
c
it
y
(
γ
)
:
de
te
r
m
in
e
s
t
he
e
ll
ip
ti
c
it
y of
t
he
2D
i
m
a
ge
v)
W
a
ve
le
ngt
h
(
λ
)
:
in
di
c
a
te
s
t
he
di
s
ta
nc
e
be
twe
e
n t
he
hi
gh
e
s
t
poi
n
ts
T
he
w
a
ve
le
t
s
a
r
e
a
ppl
ie
d
a
t
di
f
f
e
r
e
nt
or
ie
nt
a
ti
ons
a
nd
s
c
a
le
s
.
F
e
a
tu
r
e
s
s
uc
h
a
s
e
ne
r
gy,
c
ont
r
a
s
t,
a
nd
va
r
ia
nc
e
a
r
e
obt
a
in
e
d.
T
h
e
out
put
of
th
e
G
a
bor
w
a
ve
le
t
tr
a
ns
f
o
r
m
a
ti
on
is
a
s
e
t
of
f
e
a
tu
r
e
ve
c
to
r
s
us
e
d
in
th
e
c
lu
s
te
r
in
g
s
te
p
f
or
de
f
e
c
t
de
te
c
ti
on.
B
e
c
a
u
s
e
of
th
e
bi
ol
ogi
c
a
l
r
e
le
va
nc
e
of
G
a
bor
w
a
ve
s
,
th
e
y
a
r
e
us
e
d
f
r
e
que
nt
ly
, a
nd t
he
ir
m
a
th
e
m
a
ti
c
a
l
pr
ope
r
ti
e
s
a
r
e
de
f
in
e
d a
s
i
n (
1)
:
,
(
)
=
,
2
2
e
x
p
(
−
,
2
2
2
2
)
[
e
x
p
(
,
)
−
e
x
p
(
−
2
2
)
]
(
1)
W
he
r
e
th
e
or
ie
nt
a
ti
on
of
th
e
G
a
bor
w
a
ve
le
t
a
nd
th
e
s
c
a
le
a
r
e
gi
ve
n
by
θ
a
nd
r
e
s
pe
c
ti
ve
ly
,
by
x
=
(
p
,
q
)
r
e
pr
e
s
e
nt
in
g t
he
s
pa
ti
a
l
dom
a
in
of
t
he
G
a
bor
w
a
ve
le
t
k
θ
,
is
t
he
w
a
ve
ve
c
to
r
a
nd i
s
gi
ve
n a
s
i
n (
2)
:
,
=
e
x
p
(
)
(
2)
w
he
r
e
=
,
=
2
,
=
√
2
.
G
a
bor
w
a
ve
le
ts
a
r
e
u
s
e
d
due
to
th
e
ir
s
im
il
a
r
it
y
to
th
e
hum
a
n
vi
s
ua
l
s
y
s
te
m
a
nd
be
c
a
u
s
e
th
e
y
don’
t
r
e
qui
r
e
de
f
e
c
t
-
f
r
e
e
s
a
m
pl
e
s
or
th
r
e
s
hol
d
s
e
tt
in
gs
[
10]
.
T
h
e
y
e
x
tr
a
c
t
im
a
ge
f
e
a
tu
r
e
s
li
ke
e
dge
s
a
nd
poi
nt
s
f
or
c
la
s
s
if
ic
a
ti
on
ba
s
e
d
on
s
p
a
ti
a
l,
s
pe
c
tr
a
l,
or
te
xt
ur
e
pr
ope
r
t
ie
s
,
r
e
duc
in
g
di
m
e
ns
io
na
li
ty
a
nd
r
e
m
ovi
ng
unne
c
e
s
s
a
r
y
da
ta
f
or
e
a
s
ie
r
pr
oc
e
s
s
in
g.
F
ig
ur
e
3
il
lu
s
tr
a
te
s
th
e
a
ppl
ic
a
ti
on
of
G
a
bor
w
a
ve
le
ts
to
a
2D
im
a
ge
.
F
ig
ur
e
3(
a
)
de
pi
c
ts
th
e
r
e
a
l
w
a
ve
le
ts
,
w
hi
le
F
ig
ur
e
3(
b)
pr
e
s
e
nt
s
th
e
im
a
gi
na
r
y
w
a
ve
le
ts
a
t
or
ie
nt
a
ti
ons
of
0,
π
/4
, π
/2
, a
nd 3π
/4
r
a
di
a
ns
, a
nd s
c
a
le
s
r
a
ngi
ng f
r
om
0.2 to 0.6.
F
ig
ur
e
2. B
lo
c
k di
a
gr
a
m
of
t
he
pr
opos
e
d m
e
th
od
(
a
)
(
b)
F
ig
ur
e
3. G
a
bor
w
a
ve
le
ts
f
or
a
2D
i
m
a
ge
of
(
a
)
r
e
a
l
w
a
ve
le
ts
a
nd
(
b)
i
m
a
gi
na
r
y w
a
ve
le
ts
2.2
.
K
-
m
e
an
s
c
lu
s
t
e
r
in
g
K
-
m
e
a
ns
c
lu
s
te
r
in
g
is
a
n
uns
upe
r
vi
s
e
d
a
lg
or
it
hm
th
a
t
di
vi
de
s
da
ta
in
to
s
ubgr
oups
c
a
ll
e
d
c
lu
s
te
r
s
ba
s
e
d
on
c
ha
r
a
c
te
r
is
ti
c
s
li
ke
s
iz
e
,
s
h
a
pe
,
or
ie
nt
a
ti
on,
a
nd
s
c
a
le
.
T
he
num
be
r
of
c
lu
s
te
r
s
(
K
)
is
de
te
r
m
in
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 5, O
c
to
be
r
2025
:
4279
-
4289
4282
us
in
g
th
e
E
lb
ow
m
e
th
od.
K
-
m
e
a
ns
m
in
im
iz
e
s
th
e
s
um
of
di
s
ta
nc
e
s
be
tw
e
e
n
da
ta
poi
nt
s
a
nd
th
e
ir
ne
a
r
e
s
t
c
e
nt
r
oi
ds
,
gr
oupi
ng
s
im
il
a
r
da
ta
poi
nt
s
in
to
c
lu
s
te
r
s
w
it
hout
r
e
qui
r
in
g
tr
a
in
in
g.
T
he
pr
oc
e
s
s
r
e
pe
a
ts
unt
il
no
f
ur
th
e
r
c
lu
s
te
r
in
g i
s
pos
s
ib
le
, e
ns
ur
in
g t
ha
t
e
a
c
h da
ta
s
e
t
be
lo
ngs
t
o t
he
c
lu
s
te
r
w
it
h t
he
ne
a
r
e
s
t
c
e
nt
r
oi
d.
2.3
.
D
e
s
c
r
ip
t
io
n
of
t
h
e
al
gor
it
h
m
I
n
K
-
m
e
a
ns
c
lu
s
te
r
in
g,
th
e
va
lu
e
of
K
in
di
c
a
te
s
th
e
num
be
r
of
c
lu
s
te
r
s
.
F
or
in
s
ta
nc
e
,
if
th
e
va
lu
e
of
K
is
s
e
t
to
2,
two
c
lu
s
te
r
s
w
il
l
be
c
r
e
a
te
d;
if
it
is
s
e
t
to
3,
th
r
e
e
c
lu
s
te
r
s
w
il
l
be
c
r
e
a
te
d.
T
he
f
or
m
a
ti
on
of
K
c
lu
s
te
r
s
e
ns
ur
e
s
th
a
t
a
d
a
ta
s
e
t
is
di
vi
de
d
in
to
gr
oups
ba
s
e
d
on
th
e
s
im
il
a
r
it
y
of
a
tt
r
ib
ut
e
s
.
T
he
K
-
m
e
a
ns
c
lu
s
te
r
in
g a
lg
or
it
hm
ope
r
a
te
s
a
s
s
how
n i
n A
lg
or
it
hm
1.
A
lg
or
it
hm
1. K
-
m
e
a
ns
c
lu
s
te
r
in
g
1)
C
hoos
e
t
he
numbe
r
of
K
c
lu
s
te
r
s
you wa
nt
t
o c
r
e
a
te
.
2)
I
ni
ti
a
li
z
e
K
c
e
nt
r
oi
ds
r
a
ndoml
y f
r
om
t
he
da
ta
poi
nt
s
.
3)
A
s
s
ig
n
e
a
c
h
pi
xe
l
to
th
e
c
lo
s
e
s
t
c
e
nt
r
oi
d
on
th
e
ba
s
is
of
it
s
E
uc
li
de
a
n
di
s
ta
nc
e
.
T
he
E
uc
li
de
a
n
di
s
ta
n
c
e
is
c
a
lc
ul
a
te
d u
s
in
g t
he
f
or
m
ul
a
gi
ve
n i
n (
3)
.
Euc
l
id
e
a
n
d
is
t
a
n
c
e
=
√
∑
(
x
i
−
y
i
n
i
=
1
)
2
(
3)
4)
R
e
c
a
lc
ul
a
te
t
h
e
c
e
nt
r
oi
d a
s
t
he
m
e
a
n of
a
ll
pi
xe
ls
a
s
s
ig
ne
d t
o i
t.
5)
R
e
pe
a
t
s
t
e
ps
3 a
nd 4 unti
l
c
onve
r
ge
n
c
e
(
w
he
n t
he
c
e
nt
r
oi
d pos
it
io
ns
s
to
p c
ha
ngi
ng s
ig
ni
f
ic
a
nt
ly
)
.
6)
A
s
s
ig
n e
a
c
h pi
xe
l
to
t
he
c
lu
s
te
r
t
o w
hi
c
h i
ts
c
e
nt
r
oi
d be
lo
ngs
.
7)
Re
-
s
ha
pe
t
he
c
lu
s
te
r
e
d da
ta
b
a
c
k i
nt
o t
he
or
ig
in
a
l
im
a
ge
s
ha
pe
.
8)
O
pt
i
on
a
l
ly
,
a
ppl
y
po
s
t
-
pr
o
c
e
s
s
i
ng
t
o
t
h
e
s
e
g
m
e
nt
e
d
im
a
g
e
(
e
.
g.,
s
m
oot
hi
n
g
or
m
or
p
hol
ogi
c
a
l
op
e
r
a
ti
on
s
)
.
9)
U
s
e
t
he
s
e
gm
e
nt
e
d i
m
a
ge
t
o de
te
c
t
a
nd c
la
s
s
if
y de
f
e
c
ts
i
n t
he
vi
a
l.
I
n
th
is
pr
oc
e
s
s
,
th
e
in
put
s
a
m
pl
e
is
de
f
in
e
d
a
s
S
=
{
p
1
,
p
2
,
.
.
.
,
p
m
}
.
T
h
e
va
r
ia
bl
e
K
de
not
e
s
th
e
num
be
r
of
c
lu
s
te
r
s
,
a
nd
N
s
ig
ni
f
ie
s
th
e
m
a
xi
m
um
num
be
r
of
r
e
pe
ti
ti
ons
a
ll
ow
e
d.
T
he
out
put
is
d
e
not
e
d
by
C
=
{
C
1
,
C
2
,
.
.
.
,
C
K
}
. T
he
pr
oc
e
dur
e
i
nvol
ve
s
t
he
f
ol
lo
w
in
g s
t
e
ps
:
i)
F
r
om
s
a
m
pl
e
S
,
s
e
le
c
t
r
a
ndoml
y
k
num
be
r
s
of
s
a
m
pl
e
s
w
hi
c
h
a
r
e
in
it
ia
l
c
e
nt
e
r
ve
c
to
r
s
of
k
c
lu
s
te
r
s
r
e
pr
e
s
e
nt
e
d a
s
:
{
μ
1
,
μ
2
,
.
.
.
,
μ
k
}
.
ii)
F
or
t
he
numbe
r
n=
1, 2, ..., N, a
ppl
y t
he
f
ol
lo
w
in
g.
a)
C
i
s
i
ni
ti
a
li
z
e
d a
s
C
t
=
∅
ii
j
N
t
=
1
,
2
,
…
.
k
w
he
r
e
C
i
s
c
la
s
s
di
vi
s
io
n
b)
D
is
ta
nc
e
be
tw
e
e
n
p
i
(
f
or
a
ll
i=
1,
2,
...,
m
)
a
nd
e
ve
r
y
c
lu
s
te
r
c
e
nt
e
r
μ
j
(
f
or
a
ll
j=
1,
2,
...,
k)
is
c
a
lc
ul
a
te
d
a
s
i
n (
4)
a
nd (
5)
.
s
ij
=
p
i
−
μ
j2
2
(
4)
C
α
i
=
C
α
i
∪
{
p
i
}
(
5)
F
or
th
e
s
a
m
pl
e
p
i
ha
vi
ng
th
e
s
m
a
ll
e
s
t
di
s
ta
nc
e
f
r
om
μ
j
is
te
r
m
e
d
a
s
s
ij
a
nd
it
s
c
a
t
e
gor
y
is
r
e
pr
e
s
e
nt
e
d
a
s
α
i
, s
o t
ha
t
th
e
out
put
c
a
te
gor
y i
s
upda
te
d
a
s
i
n (
6)
:
C
α
i
=
C
α
i
∪
{
p
i
}
,
μ
j
=
1
|
C
j
|
∑
p
p
ϵ
C
j
(
6)
c)
F
or
th
e
out
put
c
la
s
s
C
j
,
di
vi
de
a
ll
th
os
e
s
a
m
pl
e
poi
nt
s
by
(
4)
f
or
a
ll
j=
1,
2,
3....,
k
to
obt
a
in
th
e
ne
w
c
lu
s
te
r
c
e
nt
e
r
s
a
s
i
n (
7)
.
μ
j
=
1
|
C
j
|
∑
p
p
ϵ
C
j
(
7)
d)
R
e
pe
a
t
s
t
e
p (
c
)
f
or
a
ll
k s
a
m
pl
e
s
.
N
ow
f
or
p
i
th
e
s
m
a
ll
e
s
t
di
s
ta
nc
e
f
r
om
th
e
c
e
nt
e
r
of
th
e
c
lu
s
te
r
c
e
nt
e
r
is
m
a
r
ke
d
a
s
s
ij
w
hos
e
c
a
te
gor
y
is
gi
ve
n by
α
i
, w
it
h t
hi
s
upda
te
out
put
a
s
i
n (
8)
:
C
α
i
=
C
α
i
∪
{
p
i
}
μ
j
=
1
|
C
j
|
∑
p
p
ϵ
C
j
(
8)
iii)
T
he
pr
oc
e
dur
e
is
r
e
pe
a
te
d
f
or
a
ll
th
e
v
a
lu
e
s
of
j=
1,
2,
3....,
k
f
or
a
ll
th
e
va
lu
e
s
of
th
e
k
s
a
m
pl
e
s
,
th
e
c
e
nt
e
r
ve
c
to
r
s
of
th
e
gr
oup
ha
ve
not
c
ha
nge
d,
a
nd
th
e
f
in
a
l
c
la
s
s
di
vi
s
io
n
out
put
C
=
{
C
1
,
C
2
,
.
.
.
,
C
K
}
is
obt
a
in
e
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
ut
om
at
e
d v
ia
l
de
fe
c
t
in
s
pe
c
ti
on us
in
g
G
abo
r
w
av
e
l
e
ts
and
k
-
m
e
ans
c
lu
s
te
r
in
g
(
V
is
hw
anat
ha C
. R
.)
4283
T
he
s
u
c
c
e
s
s
of
th
e
K
-
m
e
a
n
s
a
lg
or
it
hm
de
pe
nds
on
th
e
va
lu
e
of
k
th
a
t
w
e
c
hoo
s
e
.
T
h
e
r
e
a
r
e
va
r
io
us
w
a
ys
to
de
r
iv
e
a
n
opt
im
a
l
num
be
r
of
c
lu
s
te
r
s
.
O
ne
s
uc
h
w
a
y
is
to
us
e
th
e
E
lb
ow
m
e
th
od.
T
hi
s
m
e
th
od
us
e
s
w
it
hi
n
-
c
lu
s
te
r
s
um
of
s
qua
r
e
s
(
W
C
S
S
)
t
o ge
t
th
e
opt
im
a
l
num
be
r
of
c
lu
s
te
r
s
.
W
C
S
S
is
de
f
in
e
d a
s
i
n (
9)
:
∑
P
i
in
C
l
u
s
te
r
1
d
is
t
a
n
c
e
(
P
i
C
1
)
2
+
∑
P
i
in
C
l
u
s
te
r
2
d
is
t
a
n
c
e
(
P
i
C
2
)
2
+
∑
P
i
in
C
l
u
s
te
r
3
d
is
t
a
n
c
e
(
P
i
C
3
)
2
(
9)
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
D
a
ta
s
e
t
de
s
c
r
ip
ti
on,
th
e
s
tu
dy
us
e
s
a
c
us
to
m
-
c
ur
a
te
d
da
ta
s
e
t
s
pe
c
if
ic
a
ll
y
de
s
ig
ne
d
f
or
vi
a
l
de
f
e
c
t
in
s
pe
c
ti
on,
a
ddr
e
s
s
in
g
li
m
it
a
ti
ons
of
publ
ic
ly
a
va
il
a
bl
e
da
ta
s
e
ts
,
a
s
not
e
d
by
H
u
e
t
al
.
[
24
]
.
T
hi
s
ta
il
or
e
d
da
ta
s
e
t
e
nha
nc
e
s
de
f
e
c
t
de
te
c
ti
on
a
c
c
ur
a
c
y
by
c
lo
s
e
ly
r
e
f
le
c
ti
ng
r
e
a
l
-
w
or
ld
s
c
e
na
r
io
s
.
A
ugm
e
nt
a
ti
on
te
c
hni
que
s
li
ke
r
ot
a
ti
on
a
nd
s
c
a
li
ng
e
xpa
nde
d
th
e
da
ta
s
e
t,
w
hi
le
r
ig
or
ous
va
li
da
ti
on
e
ns
ur
e
d
r
obus
tn
e
s
s
w
it
hout
t
r
a
di
ti
ona
l
tr
a
in
in
g.
T
he
im
a
ge
a
c
qui
s
it
io
n
s
e
tu
p
a
s
s
how
n
in
F
ig
ur
e
4
in
c
lu
de
s
a
N
e
bul
a
L
E
D
ba
c
kl
ig
ht
(
N
e
bul
a
-
80R
)
f
or
a
dj
us
ta
bl
e
br
ig
ht
ne
s
s
,
a
T
a
m
r
on
m
a
c
hi
n
e
vi
s
io
n
c
a
m
e
r
a
(
5
M
P
,
V
C
X
G
-
52M
R
)
w
it
h
a
50
m
m
B
a
um
e
r
le
ns
(
M
118F
M
50)
,
a
nd
a
c
onve
yor
s
ys
te
m
.
F
ig
ur
e
4(
a
)
s
how
s
th
e
c
a
m
e
r
a
pos
it
io
ne
d
to
c
a
pt
ur
e
th
e
vi
a
l
s
ur
f
a
c
e
r
e
gi
on, while
F
ig
ur
e
4(
b
)
i
ll
us
tr
a
te
s
t
he
c
a
m
e
r
a
s
e
tu
p f
o
r
c
a
pt
ur
in
g t
he
vi
a
l
bo
tt
om
r
e
gi
on.
F
ig
ur
e
4
(
c
)
de
pi
c
ts
th
e
vi
a
l
pl
a
c
e
d
w
it
h
th
e
ba
c
kl
ig
ht
f
or
im
a
ge
a
c
qui
s
it
io
n.
T
he
c
om
put
e
r
f
e
a
tu
r
e
s
a
2.8
G
H
z
5t
h
ge
n
c
or
e
i5
pr
oc
e
s
s
or
,
8
G
B
R
A
M
,
N
vi
di
a
R
T
X
3070
G
P
U
,
a
n
d
1
T
B
s
to
r
a
ge
.
T
hi
s
c
onf
ig
ur
a
ti
on
s
uppor
ts
pr
e
c
is
e
i
m
a
ge
c
a
pt
ur
e
a
nd e
f
f
e
c
ti
ve
de
f
e
c
t
de
te
c
ti
on.
(
a
)
(
b)
(
c
)
F
ig
ur
e
4. S
ys
te
m
s
e
tu
p f
or
i
m
a
ge
a
c
qui
s
it
io
n
of
(
a
)
ca
m
e
r
a
pos
i
ti
one
d f
or
c
a
pt
ur
in
g vi
a
l
s
ur
f
a
c
e
r
e
gi
on
,
(
b)
c
a
m
e
r
a
po
s
i
ti
o
ne
d f
or
c
a
pt
ur
in
g v
ia
l
b
ot
t
om
r
e
gi
o
n
,
a
nd (
c
)
v
i
a
l
pl
a
c
e
d
w
it
h
b
a
c
kl
i
ght
f
or
i
m
a
g
e
a
c
q
ui
s
it
io
n
T
he
s
tu
dy
us
e
s
vi
a
ls
w
it
h
de
f
e
c
ts
s
uc
h
a
s
bl
a
c
k
s
pot
s
,
dus
t,
s
c
r
a
tc
he
s
,
d
e
nt
s
,
bubbl
e
s
,
m
out
h
c
r
a
c
ks
,
ne
c
k
c
r
a
c
ks
,
a
nd
m
out
h
a
nd
r
im
c
hi
ppi
ng.
M
os
t
of
th
e
s
e
de
f
e
c
ts
a
r
e
c
a
us
e
d
dur
in
g
th
e
vi
a
l
m
a
nuf
a
c
tu
r
in
g
pr
oc
e
s
s
. T
o de
m
ons
tr
a
te
t
he
a
lg
or
it
hm
, a
s
a
m
pl
e
vi
a
l
im
a
ge
w
it
h a
m
out
h c
r
a
c
k a
s
a
de
f
e
c
t
is
t
a
ke
n, a
s
s
how
n
in
F
ig
ur
e
5.
T
he
r
e
s
ul
ta
nt
im
a
ge
s
of
th
e
vi
a
l
a
f
te
r
th
e
a
ppl
i
c
a
ti
on
of
G
a
bor
w
a
ve
le
ts
w
it
h
a
k
e
r
ne
l
s
iz
e
of
256
×
256
a
r
e
s
how
n
in
F
ig
ur
e
6.
F
ig
ur
e
6(
a
)
de
pi
c
ts
th
e
r
e
a
l
w
a
ve
le
t
r
e
s
pons
e
,
w
hi
le
F
ig
ur
e
6(
b)
s
how
s
th
e
im
a
gi
na
r
y
w
a
ve
le
t
r
e
s
pons
e
.
I
n
bot
h
s
ubf
ig
ur
e
s
,
th
e
o
r
ie
nt
a
ti
on
s
a
r
e
0,
π
/4
,
π
/2
,
a
nd
3π
/4
r
a
di
a
ns
f
r
om
le
f
t
to
r
ig
ht
a
nd t
he
di
f
f
e
r
e
nt
s
c
a
le
s
a
r
e
0.2, 0.3, 0.4, 0.5, a
nd 0.6 f
r
om
t
op t
o bott
om
.
T
he
E
lb
ow
m
e
th
od
in
vol
ve
s
e
xe
c
ut
in
g
K
-
m
e
a
ns
c
lu
s
te
r
in
g
on
th
e
da
ta
s
e
t
f
or
K
va
lu
e
s
r
a
ngi
ng
f
r
om
1
to
10,
f
ol
lo
w
e
d
by
c
a
lc
ul
a
ti
ng
th
e
W
C
S
S
f
or
e
a
c
h
K
va
lu
e
.
A
c
ur
ve
is
th
e
n
pl
ot
te
d
be
twe
e
n
th
e
c
a
lc
ul
a
te
d
W
C
S
S
va
lu
e
s
a
nd
th
e
c
or
r
e
s
ponding
K
va
lu
e
s
.
T
he
opt
im
a
l
K
f
or
c
lu
s
te
r
in
g
is
de
te
r
m
in
e
d
by
id
e
nt
if
yi
ng
th
e
s
ha
r
p
be
nd,
or
"
E
lb
ow
,"
in
th
e
pl
ot
te
d
c
ur
ve
,
w
hi
c
h
in
di
c
a
te
s
th
e
poi
nt
a
t
w
hi
c
h
a
ddi
ng
m
or
e
c
lu
s
te
r
s
doe
s
not
s
ig
ni
f
ic
a
nt
ly
i
m
pr
ove
t
he
m
ode
l'
s
pe
r
f
or
m
a
nc
e
.
T
he
im
a
ge
w
it
h
s
c
r
a
tc
he
s
in
th
e
bot
to
m
(
or
ba
s
e
)
r
e
gi
on
u
nde
r
goe
s
noi
s
e
r
e
m
ova
l
a
nd
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
us
in
g
G
a
bor
w
a
ve
le
ts
,
a
s
il
lu
s
tr
a
te
d
in
F
ig
ur
e
7.
F
ig
ur
e
7(
a
)
s
how
s
th
e
in
put
im
a
ge
w
it
h
s
c
r
a
t
c
he
s
in
t
he
bot
to
m
r
e
gi
on. F
ig
ur
e
7(
b
)
pr
e
s
e
nt
s
t
he
pr
oc
e
s
s
e
d i
m
a
ge
a
f
te
r
f
e
a
tu
r
e
e
xt
r
a
c
ti
on, whe
r
e
f
e
a
tu
r
e
s
s
uc
h a
s
e
ne
r
gy,
c
ont
r
a
s
t,
a
nd
va
r
ia
nc
e
a
r
e
e
xt
r
a
c
te
d.
A
G
a
bor
w
a
ve
le
t
ke
r
ne
l
s
iz
e
of
32
c
a
pt
ur
e
s
f
e
a
tu
r
e
s
a
t
m
ul
ti
pl
e
s
c
a
le
s
,
w
hi
le
16
or
ie
nt
a
ti
ons
e
nha
nc
e
de
f
e
c
t
de
te
c
ti
on
f
r
om
va
r
io
us
a
ngl
e
s
.
T
he
im
a
ge
is
th
e
n
s
e
gm
e
nt
e
d
us
in
g
K
-
m
e
a
ns
c
lu
s
te
r
in
g
w
it
h
k=
3
c
lu
s
te
r
s
,
a
s
s
how
n
in
F
ig
u
r
e
7(
c
)
.
T
hi
s
pa
r
ti
ti
oni
ng
di
s
ti
ngui
s
he
s
a
c
tu
a
l
de
f
e
c
ts
f
r
om
noi
s
e
.
A
lt
hough
in
c
r
e
a
s
in
g
th
e
num
be
r
of
c
lu
s
te
r
s
c
oul
d
im
pr
ove
s
e
gm
e
nt
a
ti
on,
it
a
ls
o
r
a
i
s
e
s
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
. T
he
r
e
f
or
e
, s
e
le
c
ti
ng a
n opti
m
a
l
k va
lu
e
i
s
c
r
uc
ia
l
f
or
ba
la
nc
in
g e
f
f
ic
ie
nc
y a
nd da
ta
va
r
ia
ti
on
c
a
pt
ur
e
.
F
in
a
ll
y,
th
e
C
a
nny
e
dge
ope
r
a
to
r
is
a
ppl
ie
d
to
hi
ghl
ig
ht
de
f
e
c
ts
in
th
e
s
e
gm
e
nt
e
d
vi
a
l
im
a
ge
, a
s
de
pi
c
te
d i
n F
ig
ur
e
7(
d)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 5, O
c
to
be
r
2025
:
4279
-
4289
4284
F
ig
ur
e
5. A
s
a
m
pl
e
vi
a
l
im
a
ge
w
it
h a
m
out
h c
r
a
c
k de
f
e
c
t
(
a
)
(
b)
F
ig
ur
e
6. R
e
s
ul
ti
ng via
l
im
a
ge
s
a
f
te
r
a
ppl
yi
ng G
a
bor
w
a
ve
le
ts
of
(
a
)
r
e
a
l
a
nd
(
b)
i
m
a
gi
na
r
y
(
a
)
(
b)
(
c
)
(
d)
F
ig
ur
e
7. S
te
pw
is
e
pr
oc
e
s
s
in
g of
a
vi
a
l
im
a
ge
w
it
h
s
c
r
a
tc
he
s
i
n t
he
bot
to
m
r
e
gi
on
of
(
a
)
i
nput
i
m
a
ge
w
it
h
s
c
r
a
tc
he
s
i
n t
he
bot
to
m
r
e
gi
on
,
(
b)
d
e
noi
s
in
g a
nd f
e
a
tu
r
e
e
xt
r
a
c
t
io
n us
in
g G
a
bor
w
a
ve
le
t
,
(
c
)
s
e
gm
e
nt
e
d
im
a
ge
us
in
g K
-
m
e
a
ns
, a
nd
(
d)
i
de
nt
if
ie
d c
lu
s
te
r
e
dge
s
i
n t
he
ou
tp
ut
i
m
a
ge
T
he
pr
opos
e
d
m
e
th
od
in
s
pe
c
t
s
v
a
r
io
us
de
f
e
c
t
s
c
om
m
onl
y
f
ound
in
pha
r
m
a
c
e
ut
ic
a
l
vi
a
l
s
to
e
ns
ur
e
c
om
pr
e
he
ns
iv
e
qua
li
ty
a
s
s
e
s
s
m
e
nt
.
A
to
ta
l
of
254
de
f
e
c
ti
ve
vi
a
ls
w
e
r
e
a
na
ly
z
e
d
,
e
nc
om
pa
s
s
in
g
a
di
ve
r
s
e
r
a
nge
of
de
f
e
c
ts
,
in
c
lu
di
ng
bl
a
c
k
s
pot
s
(
35
oc
c
ur
r
e
nc
e
s
)
,
c
r
a
c
ks
(
60
oc
c
ur
r
e
nc
e
s
)
,
s
c
r
a
tc
he
s
(
80
oc
c
ur
r
e
nc
e
s
)
,
c
hi
ppi
ngs
(
65
oc
c
ur
r
e
nc
e
s
)
,
a
s
w
e
ll
a
s
ot
he
r
le
s
s
f
r
e
que
nt
d
e
f
e
c
ts
s
uc
h
a
s
bubble
s
(
10
o
c
c
ur
r
e
nc
e
s
)
a
nd
m
is
c
e
ll
a
ne
ous
ty
pe
s
li
ke
f
la
t
s
li
m
r
in
gs
,
de
nt
s
,
a
nd
ot
he
r
s
(
4
oc
c
ur
r
e
nc
e
s
)
.
T
hi
s
da
ta
s
e
t
a
ll
ow
s
f
or
a
th
or
ough
e
xa
m
in
a
ti
on of
de
f
e
c
t
de
te
c
ti
on c
a
pa
bi
li
ti
e
s
a
c
r
os
s
di
f
f
e
r
e
nt
c
a
te
gor
ie
s
.
T
he
a
lg
or
it
hm
s
im
pl
e
m
e
nt
e
d
e
xhi
bi
t
va
r
yi
ng
ti
m
e
a
nd
s
pa
c
e
c
om
pl
e
xi
ti
e
s
.
C
a
nny
e
dge
de
te
c
ti
on
ope
r
a
te
s
in
(
)
ti
m
e
,
w
he
r
e
is
th
e
num
be
r
of
pi
xe
l
s
,
w
hi
le
th
e
G
a
bor
w
a
ve
le
t
a
nd
K
-
m
e
a
ns
c
lu
s
te
r
in
g
e
xhi
bi
t
ti
m
e
c
om
pl
e
xi
ti
e
s
of
(
.
2
.
)
a
nd
(
.
.
.
)
r
e
s
pe
c
ti
v
e
ly
,
w
he
r
e
r
e
pr
e
s
e
nt
s
th
e
num
be
r
of
G
a
bor
f
il
te
r
s
,
de
not
e
s
th
e
num
be
r
of
it
e
r
a
ti
ons
,
a
nd
s
ig
ni
f
ie
s
th
e
num
be
r
of
di
m
e
ns
io
ns
.
T
he
s
pa
c
e
c
om
pl
e
xi
ti
e
s
f
or
m
os
t
ope
r
a
ti
ons
a
r
e
(
)
,
e
xc
e
pt
f
or
K
-
m
e
a
n
s
c
lu
s
te
r
in
g,
w
hi
c
h
ha
s
a
s
pa
c
e
c
om
pl
e
xi
ty
of
(
.
)
,
w
he
r
e
is
th
e
num
be
r
of
di
m
e
ns
io
ns
.
T
he
pe
r
f
or
m
a
nc
e
of
t
he
im
pl
e
m
e
nt
e
d
s
ys
te
m
is
e
va
lu
a
te
d
us
in
g
a
c
onf
us
io
n
m
a
tr
ix
,
w
hi
c
h
is
one
of
th
e
popula
r
m
e
a
s
ur
e
s
us
e
d
f
or
th
e
a
n
a
ly
s
is
of
pe
r
f
or
m
a
nc
e
pa
r
a
m
e
te
r
s
s
u
c
h
a
s
r
e
c
a
ll
,
pr
e
c
is
io
n,
a
c
c
ur
a
c
y,
a
nd
F
1
-
s
c
or
e
.
P
e
r
f
or
m
a
nc
e
pa
r
a
m
e
te
r
s
a
r
e
c
om
put
e
d
u
s
in
g
te
r
m
in
ol
ogi
e
s
s
uc
h
a
s
tr
ue
pos
it
iv
e
(
TP
)
,
t
r
ue
ne
ga
ti
ve
(
TN
)
,
f
a
ls
e
pos
it
iv
e
(
FP
)
,
a
nd
f
a
ls
e
ne
ga
ti
ve
(
FN
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
ut
om
at
e
d v
ia
l
de
fe
c
t
in
s
pe
c
ti
on us
in
g
G
abo
r
w
av
e
l
e
ts
and
k
-
m
e
ans
c
lu
s
te
r
in
g
(
V
is
hw
anat
ha C
. R
.)
4285
F
ig
ur
e
8
pr
e
s
e
nt
s
th
e
c
onf
us
io
n
m
a
tr
ix
s
um
m
a
r
iz
in
g
th
e
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
. T
he
r
e
c
a
ll
,
pr
e
c
is
io
n,
a
c
c
ur
a
c
y,
a
nd F
1
-
s
c
or
e
c
a
n be
c
a
lc
ul
a
te
d a
s
i
n (
10)
t
o (
15)
.
=
+
(
10)
=
+
(
11)
=
+
(
12)
1
−
=
2
×
∗
+
(
13)
F
ig
ur
e
8. C
onf
us
io
n m
a
tr
ix
of
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
T
he
a
r
e
a
unde
r
th
e
c
ur
ve
-
r
e
c
e
iv
e
r
ope
r
a
ti
ng
c
h
a
r
a
c
te
r
is
ti
c
(
AUC
-
R
O
C
)
c
ur
ve
a
nd
th
e
AUC
th
e
pr
e
c
is
io
n
-
r
e
c
a
ll
(
P
R
)
c
ur
ve
a
r
e
c
om
m
on
m
e
tr
ic
s
us
e
d
to
e
v
a
lu
a
te
th
e
p
e
r
f
or
m
a
nc
e
of
bi
na
r
y
c
la
s
s
if
ic
a
ti
on
m
ode
ls
a
s
gi
ve
n i
n (
14)
a
nd (
15)
.
_
=
∫
(
−
1
(
)
)
1
0
(
14)
_
=
∫
(
)
1
0
∙
(
)
(
15)
W
he
r
e
K
r
e
pr
e
s
e
nt
s
th
e
num
be
r
of
c
lu
s
te
r
s
.
T
h
e
A
U
C
-
R
O
C
i
s
c
a
lc
ul
a
t
e
d
c
on
s
id
e
r
in
g
th
e
TP
r
a
te
(
s
e
ns
it
iv
it
y)
a
nd
th
e
FP
r
a
te
.
T
he
TP
r
a
te
w
ou
ld
r
e
pr
e
s
e
nt
th
e
pr
opor
ti
on
of
c
or
r
e
c
tl
y
id
e
n
ti
f
ie
d
de
f
e
c
ti
v
e
vi
a
l
s
,
w
he
r
e
a
s
th
e
FP
r
a
te
w
o
ul
d
r
e
pr
e
s
e
nt
th
e
pr
opor
t
io
n
of
no
n
-
de
f
e
c
ti
v
e
vi
a
ls
i
n
c
or
r
e
c
tl
y
id
e
nt
if
ie
d a
s
de
f
e
c
ti
v
e
. T
he
A
U
C
-
P
R
m
e
a
s
ur
e
s
t
h
e
ba
l
a
nc
e
b
e
tw
e
e
n pr
e
c
i
s
io
n
a
nd r
e
c
a
ll
. P
r
e
c
i
s
io
n w
o
ul
d r
e
pr
e
s
e
nt
t
h
e
a
c
c
ur
a
c
y of
th
e
i
de
nt
if
i
c
a
ti
on
of
th
e
de
f
e
c
t
a
m
on
g
a
l
l
th
e
vi
a
ls
id
e
nt
if
ie
d
a
s
de
f
e
c
ti
v
e
,
w
hi
le
r
e
c
a
ll
(
or
s
e
n
s
it
iv
i
ty
)
w
o
ul
d
r
e
pr
e
s
e
n
t
th
e
pr
opor
ti
on
of
a
c
tu
a
l
d
e
f
e
c
ti
ve
v
ia
l
s
c
or
r
e
c
tl
y
id
e
nt
if
ie
d.
T
he
p
r
opos
e
d
m
e
th
od
ha
s
pr
odu
c
e
d
9
8.81%
r
e
c
a
ll
,
98.03%
pr
e
c
i
s
io
n,
9
8.28%
a
c
c
ur
a
c
y,
a
nd
th
e
F
1
-
s
c
or
e
obt
a
in
e
d
i
s
98.42%
.
T
he
A
U
C
-
R
O
C
a
n
d
A
U
C
-
P
R
va
lu
e
s
o
bt
a
in
e
d
a
r
e
9
8.30
a
nd
98.9
6%
, r
e
s
pe
c
ti
v
e
ly
,
w
he
n t
he
v
a
lu
e
of
k i
s
s
e
t
t
o 3
a
nd i
s
s
ho
w
n i
n F
i
gur
e
9.
T
he
pe
r
f
or
m
a
nc
e
of
th
e
pr
opos
e
d
m
e
th
odol
ogy
a
c
r
o
s
s
v
a
r
io
us
s
e
gm
e
nt
a
ti
on
s
e
tt
in
gs
is
a
s
s
e
s
s
e
d
by
in
te
gr
a
ti
ng
di
f
f
e
r
e
nt
k
va
lu
e
s
,
a
s
s
how
n
in
F
ig
ur
e
10.
F
ig
ur
e
10(
a
)
pr
e
s
e
nt
s
th
e
R
O
C
c
ur
ve
,
w
hi
le
F
ig
ur
e
10(
b)
il
lu
s
tr
a
te
s
th
e
P
R
c
ur
ve
f
or
di
f
f
e
r
e
nt
k
va
lu
e
s
.
T
he
a
n
a
ly
s
is
r
e
v
e
a
ls
th
a
t
va
r
yi
ng
K
im
pa
c
t
s
m
ode
l
pe
r
f
or
m
a
nc
e
,
w
it
h
K
=
3
yi
e
ld
in
g
op
ti
m
a
l
s
e
ns
it
iv
it
y,
s
pe
c
if
ic
it
y,
pr
e
c
is
io
n,
a
nd
r
e
c
a
ll
.
T
hi
s
in
di
c
a
te
s
th
a
t
de
f
e
c
t
de
te
c
ti
on
is
m
os
t
e
f
f
e
c
ti
ve
w
h
e
n
K
is
s
e
t
to
3,
hi
ghl
ig
ht
in
g
th
e
im
por
ta
nc
e
of
s
e
le
c
ti
ng
a
n
a
ppr
opr
ia
te
K
va
lu
e
f
or
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
.
T
he
pr
opos
e
d
m
e
th
od
f
or
vi
a
l
de
f
e
c
t
de
te
c
ti
on
ta
k
e
s
a
n
a
v
e
r
a
ge
c
om
put
a
ti
ona
l
ti
m
e
of
a
r
ound
355
m
ic
r
os
e
c
onds
f
or
a
ll
ty
pe
s
of
de
f
e
c
ts
,
a
nd
s
ugge
s
ts
a
r
e
la
ti
ve
ly
e
f
f
ic
ie
nt
pr
oc
e
s
s
in
g
s
pe
e
d.
T
he
c
om
pa
r
is
on
of
va
r
io
us
te
c
hni
que
s
or
m
e
th
odol
ogi
e
s
a
nd
th
e
one
us
e
d
in
th
is
s
tu
dy
is
gi
ve
n
in
T
a
bl
e
1,
w
hi
c
h
pr
ovi
de
s
e
m
pi
r
ic
a
l
e
vi
de
nc
e
t
ha
t
de
m
ons
tr
a
te
s
t
he
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
of
t
he
p
r
opos
e
d m
e
th
od c
om
pa
r
e
d t
o
a
lt
e
r
na
ti
ve
de
f
e
c
t
de
te
c
ti
on
a
lg
or
it
hm
s
.
T
he
pa
pe
r
pr
e
s
e
nt
s
r
e
s
ul
ts
f
r
om
c
om
pa
r
a
ti
ve
e
xpe
r
im
e
nt
s
th
a
t
hi
ghl
ig
ht
t
he
m
e
th
od'
s
h
ig
he
r
a
c
c
ur
a
c
y. O
ve
r
a
ll
, t
he
p
r
e
f
e
r
e
nc
e
f
or
t
he
p
r
opos
e
d m
e
th
od us
in
g
G
a
bor
w
a
ve
le
t
c
lu
s
te
r
in
g
a
nd
K
-
m
e
a
n
s
c
lu
s
te
r
in
g
is
ju
s
ti
f
ie
d
by
it
s
a
bi
li
ty
to
e
f
f
e
c
ti
ve
ly
e
xt
r
a
c
t
r
e
le
v
a
nt
f
e
a
tu
r
e
s
th
a
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 5, O
c
to
be
r
2025
:
4279
-
4289
4286
di
s
ti
ngui
s
h
be
twe
e
n
de
f
e
c
ti
ve
a
nd
non
-
de
f
e
c
ti
ve
r
e
gi
ons
of
vi
a
ls
.
T
he
gr
a
ph
of
th
e
c
om
pa
r
is
on
of
a
c
c
ur
a
c
y
a
c
r
os
s
di
f
f
e
r
e
nt
m
e
th
ods
i
s
s
how
n i
n F
ig
ur
e
11.
F
ig
ur
e
9. T
he
A
U
C
-
R
O
C
a
nd A
U
C
-
P
R
c
ur
ve
s
(
a
)
(
b)
F
ig
ur
e
10. P
e
r
f
or
m
a
nc
e
e
va
lu
a
ti
on me
tr
ic
s
of
(
a
)
r
e
c
e
iv
e
r
ope
r
a
ti
ng c
ha
r
a
c
te
r
is
ti
c
c
ur
ve
a
nd
(
b)
pr
e
c
is
io
n
-
r
e
c
a
ll
c
ur
ve
F
ig
ur
e
11. C
om
pa
r
is
on of
pr
e
c
is
io
n a
c
r
os
s
di
f
f
e
r
e
nt
m
e
th
ods
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
A
ut
om
at
e
d v
ia
l
de
fe
c
t
in
s
pe
c
ti
on us
in
g
G
abo
r
w
av
e
l
e
ts
and
k
-
m
e
ans
c
lu
s
te
r
in
g
(
V
is
hw
anat
ha C
. R
.)
4287
T
a
bl
e
1.
T
he
pe
r
f
or
m
a
nc
e
of
va
r
io
us
m
e
th
ods
S
l
. N
o.
M
e
t
hodol
ogy
A
c
c
ur
a
c
y
(%)
1
M
ul
t
i
-
c
ha
nne
l
s
e
gm
e
nt
a
t
i
on
[
1]
95
2
X
-
r
a
y a
nom
a
l
y de
t
e
c
t
i
on [
2]
97.4
3
H
e
ur
i
s
t
i
c
m
e
t
hod [
3]
94.9
4
T
hr
e
s
hol
di
ng [
4]
95.6
5
H
or
i
z
ont
a
l
i
nt
e
r
c
e
pt
pr
oj
e
c
t
i
on
[
5]
98
6
F
T
A
D
S
P
a
nd W
T
M
F
[
6]
95.00
7
C
a
nny e
dge
de
t
e
c
t
i
on [
7
]
95.33
(
a
vg)
8
C
l
us
t
e
r
i
ng [
8]
94.12
9
H
A
M
V
[
9]
95.00
10
C
a
nny e
dge
de
t
e
c
t
or
[
10]
95.00
11
D
D
S
C
N
e
t
[
15]
86.9 (
hi
ghe
s
t
)
12
Y
O
L
O
v3 [
16]
97.66
13
Y
O
L
O
v5s
[
17
]
67.5
14
LF
-
Y
O
L
O
v4 [
18]
97.83
15
I
m
pr
ove
d Y
O
L
O
v3 [
19]
91.12
16
P
r
opos
e
d m
e
t
hod
98.28
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
s
tu
dy
a
dopt
s
a
nove
l
a
ppr
oa
c
h
by
f
oc
u
s
in
g
on
th
e
de
ve
lo
pm
e
nt
a
nd
e
v
a
lu
a
ti
on
of
a
n
a
lg
or
it
hm
f
or
de
f
e
c
t
de
te
c
ti
on
in
v
ia
ls
.
T
he
a
lg
or
it
hm
is
de
s
ig
ne
d
to
a
na
ly
z
e
in
di
vi
dua
l
vi
a
l
im
a
ge
s
to
id
e
nt
if
y
a
n
d
lo
c
a
li
z
e
th
e
pr
e
s
e
nc
e
of
de
f
e
c
ts
,
w
it
hout
tr
a
in
in
g
a
m
ode
l
f
or
vi
a
l
c
la
s
s
if
ic
a
ti
on.
G
a
bor
w
a
ve
le
ts
a
r
e
a
pow
e
r
f
ul
to
ol
in
im
a
ge
p
r
oc
e
s
s
in
g
th
a
t
c
a
n
be
us
e
d
to
de
te
c
t
de
f
e
c
ts
in
vi
a
ls
by
e
xt
r
a
c
ti
ng
m
e
a
ni
ngf
ul
f
e
a
tu
r
e
s
f
r
om
th
e
im
a
ge
da
ta
.
C
om
bi
ni
ng
G
a
bor
w
a
ve
le
ts
w
it
h
c
lu
s
te
r
in
g
of
K
-
m
e
a
ns
e
na
bl
e
s
a
c
c
ur
a
te
a
nd
e
f
f
ic
ie
nt
de
te
c
ti
on
of
de
f
e
c
ti
ve
vi
a
ls
,
ut
il
iz
in
g
f
e
a
tu
r
e
s
s
uc
h
a
s
e
ne
r
gy,
c
ont
r
a
s
t,
a
nd
va
r
ia
nc
e
f
or
s
upe
r
io
r
r
e
s
ul
ts
.
R
obus
t
to
va
r
ia
ti
ons
in
il
lu
m
in
a
ti
on,
o
r
ie
nt
a
ti
on,
a
nd
s
c
a
le
,
G
a
bor
w
a
ve
le
ts
e
ns
ur
e
r
e
li
a
bi
li
ty
a
c
r
os
s
di
ve
r
s
e
m
a
nuf
a
c
tu
r
in
g
e
nvi
r
onm
e
nt
s
.
U
nl
ik
e
pr
e
d
e
f
in
e
d
m
ode
ls
,
K
-
m
e
a
ns
c
lu
s
te
r
in
g
a
d
a
pt
s
to
va
r
yi
ng
c
ha
r
a
c
te
r
is
ti
c
s
,
ope
r
a
te
s
in
a
n
uns
upe
r
vi
s
e
d
m
a
nn
e
r
,
a
nd
a
c
c
o
m
m
oda
te
s
c
ha
nge
s
in
m
a
nuf
a
c
tu
r
in
g
pr
oc
e
s
s
e
s
or
pr
oduc
t
s
pe
c
if
ic
a
ti
ons
.
T
hi
s
hol
is
ti
c
a
ppr
oa
c
h
e
nh
a
nc
e
s
th
e
m
e
th
od'
s
a
bi
li
ty
to
id
e
nt
if
y
va
r
io
us
ty
pe
s
of
de
f
e
c
ts
,
in
c
lu
di
ng
s
ubt
le
a
nom
a
li
e
s
th
a
t
m
a
y
be
c
ha
ll
e
ngi
ng
t
o
de
te
c
t
us
in
g
tr
a
di
ti
ona
l
a
lg
or
it
hm
s
.
W
it
h
a
n
a
ve
r
a
ge
c
om
put
a
ti
ona
l
ti
m
e
of
355
m
ic
r
os
e
c
onds
,
th
e
pr
op
os
e
d
m
e
th
od
d
e
m
ons
tr
a
te
s
th
e
pot
e
nt
ia
l
f
or
r
e
a
l
-
ti
m
e
de
f
e
c
t
de
te
c
ti
on,
d
e
te
c
ti
ng
s
c
r
a
tc
he
s
,
de
nt
s
,
bl
a
c
k
s
p
ot
s
,
c
hi
ppi
ng,
c
r
a
c
k
s
,
a
nd
m
or
e
.
C
om
pa
r
a
ti
ve
a
na
ly
s
is
in
di
c
a
te
s
be
tt
e
r
a
c
c
ur
a
c
y
in
a
ll
r
e
gi
ons
of
th
e
vi
a
l
te
s
te
d
f
or
de
f
e
c
ts
,
hi
ghl
ig
ht
in
g
th
e
m
e
th
od'
s
pot
e
nt
ia
l
f
or
a
dopt
io
n
in
th
e
pha
r
m
a
c
e
ut
ic
a
l
in
dus
tr
y.
T
he
m
e
th
od
de
s
c
r
ib
e
d
ope
r
a
te
s
on
in
di
vi
dua
l
vi
a
l
im
a
ge
s
s
e
que
nt
ia
ll
y.
F
or
hi
gh
-
th
r
oughput
m
a
nuf
a
c
tu
r
in
g
e
nvi
r
onm
e
nt
s
,
w
he
r
e
m
ul
ti
pl
e
vi
a
ls
a
r
e
in
s
pe
c
te
d
s
im
ul
ta
ne
ous
ly
,
th
is
a
ppr
oa
c
h
m
a
y
not
be
s
c
a
la
bl
e
.
F
ut
ur
e
r
e
s
e
a
r
c
h
c
oul
d
in
ve
s
ti
ga
t
e
pa
r
a
ll
e
l
pr
oc
e
s
s
in
g
or
ba
tc
h
in
s
pe
c
ti
on
te
c
hni
que
s
to
im
pr
ove
th
r
oughput
a
nd
a
im
a
t
e
xpl
or
in
g
th
e
in
te
gr
a
ti
on
of
opt
im
iz
e
d
ML
a
lg
or
it
hm
s
,
s
uc
h
a
s
C
N
N
s
,
S
V
M
s
,
r
a
ndom
f
or
e
s
ts
,
a
nd
de
e
p
le
a
r
ni
ng
m
e
th
ods
s
uc
h
a
s
R
-
C
N
N
s
,
U
-
N
e
t,
a
nd
G
A
N
s
, t
o i
m
pr
ove
a
ut
om
a
te
d de
f
e
c
t
c
la
s
s
if
ic
a
ti
on.
A
C
K
N
O
WL
E
D
G
M
E
N
T
S
T
he
a
ut
hor
s
w
is
h
to
e
xt
e
nd
th
e
ir
s
in
c
e
r
e
gr
a
ti
tu
de
to
M
r
.
S
unda
r
e
s
h
N
a
ya
k,
S
a
a
tv
ik
S
ol
ut
io
ns
,
B
e
nga
lu
r
u
f
or
hi
s
va
lu
a
bl
e
a
s
s
is
t
a
nc
e
in
s
e
tt
in
g
up
th
e
ha
r
dw
a
r
e
c
onf
ig
ur
a
ti
on
r
e
qui
r
e
d
f
or
th
e
e
xpe
r
im
e
nt
s
.
H
is
t
e
c
hni
c
a
l
e
xpe
r
ti
s
e
a
nd s
uppor
t
in
t
he
a
r
r
a
nge
m
e
nt
s
s
ig
ni
f
ic
a
nt
ly
c
ont
r
ib
ut
e
d t
o t
he
s
uc
c
e
s
s
f
ul
c
om
pl
e
ti
on
of
t
hi
s
r
e
s
e
a
r
c
h w
or
k.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
N
o f
undi
ng i
s
obt
a
in
e
d f
or
t
hi
s
r
e
s
e
a
r
c
h.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
d
iT
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
V
is
hw
a
na
th
a
C
.
R
.
✓
✓
✓
✓
✓
✓
✓
✓
✓
V
.
A
s
ha
✓
✓
✓
✓
✓
✓
✓
✓
C
ha
nna
ba
s
a
va
✓
✓
✓
✓
✓
S
r
e
e
ka
nt
h R
a
ll
a
pa
ll
i
✓
✓
✓
✓
✓
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
. 14, No. 5, O
c
to
be
r
2025
:
4279
-
4289
4288
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
T
he
a
ut
hor
s
de
c
la
r
e
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
th
a
t
s
uppor
t
th
e
f
in
di
ngs
of
th
is
s
tu
dy
a
r
e
a
va
il
a
bl
e
f
r
om
th
e
f
ir
s
t
a
ut
ho
r
[
V
C
R
]
,
upo
n
r
e
a
s
ona
bl
e
r
e
que
s
t.
R
E
F
E
R
E
N
C
E
S
[
1]
B
.
C
he
n,
C
.
L
i
,
P
.
Y
ua
n,
Y
.
Y
a
n,
a
nd
Y
.
Y
i
n,
“
R
e
s
e
a
r
c
h
on
de
f
e
c
t
de
t
e
c
t
i
on
of
bot
t
l
e
c
a
p
i
nt
e
r
i
or
ba
s
e
d
on
l
ow
-
a
ngl
e
a
nd
l
a
r
g
e
di
ve
r
ge
nc
e
a
ngl
e
vi
s
i
on s
y
s
t
e
m
,”
P
L
oS O
N
E
, vol
. 19, no. 5
,
M
a
y, 2024, doi
:
10
.1371/
j
our
na
l
.pone
.0303744.
[
2]
J
. R
a
pc
e
w
i
c
z
a
nd
M
. M
a
l
e
s
a
, “
A
c
t
i
ve
l
e
a
r
ni
ng i
n f
e
a
t
ur
e
e
xt
r
a
c
t
i
on f
or
gl
a
s
s
-
in
-
gl
a
s
s
de
t
e
c
t
i
on,”
E
l
e
c
t
r
oni
c
s
vol
. 13, no. 11, 2024
,
doi
:
10.3390/
e
l
e
c
t
r
oni
c
s
13112049.
[
3]
M
.
E
s
hke
va
r
i
,
M
.
J
.
R
e
z
a
e
e
,
M
.
Z
a
r
i
nba
l
,
a
nd
H
.
I
z
a
dba
khs
h,
“
A
ut
om
a
t
i
c
di
m
e
ns
i
ona
l
de
f
e
c
t
de
t
e
c
t
i
on
f
or
gl
a
s
s
vi
a
l
s
ba
s
e
d
on
m
a
c
hi
ne
vi
s
i
on:
a
he
ur
i
s
t
i
c
s
e
gm
e
nt
a
t
i
on
m
e
t
hod,”
J
our
nal
of
M
anuf
ac
t
ur
i
ng
P
r
oc
e
s
s
e
s
,
vol
.
68,
pp.
973
–
989,
2021,
doi
:
10.1016/
j
.j
m
a
pr
o.2021.06.018.
[
4]
O
.
F
a
r
ha
ngi
,
E
.
S
he
i
da
e
e
,
a
nd
A.
K
i
s
a
l
a
e
i
,
“
M
a
c
hi
ne
vi
s
i
on
f
or
de
t
e
c
t
i
ng
de
f
e
c
t
s
i
n
l
i
qui
d
bot
t
l
e
s
:
a
n
i
ndus
t
r
i
a
l
a
ppl
i
c
a
t
i
on
f
or
f
ood a
nd pa
c
ka
gi
ng s
e
c
t
or
,”
C
l
oud
C
om
put
i
ng and D
at
a Sc
i
e
nc
e
, pp. 242
–
254,
2024, doi
:
10.37256/
c
c
ds
.5220244756.
[
5]
X
.
L
i
u,
Q
.
Z
hu,
Y
.
W
a
ng,
X
.
Z
hou,
K
.
L
i
,
a
nd
X
.
L
i
u,
“
M
a
c
hi
ne
vi
s
i
on
ba
s
e
d
de
f
e
c
t
de
t
e
c
t
i
on
s
ys
t
e
m
f
or
o
r
a
l
l
i
qui
d
vi
a
l
,”
2018
13t
h W
or
l
d C
ongr
e
s
s
on I
nt
e
l
l
i
ge
nt
C
ont
r
ol
and A
ut
om
at
i
on (
W
C
I
C
A
)
, pp. 945
–
950, 2018, doi
:
10.1109/
W
C
I
C
A
.2018.8630441.
[
6]
X
.
Z
hou
e
t
al
.
,
“
A
s
ur
f
a
c
e
de
f
e
c
t
de
t
e
c
t
i
on
f
r
a
m
e
w
or
k
f
or
gl
a
s
s
bot
t
l
e
b
ot
t
om
us
i
ng
vi
s
ua
l
a
t
t
e
nt
i
on
m
ode
l
a
nd
w
a
ve
l
e
t
t
r
a
ns
f
or
m
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on I
ndus
t
r
i
al
I
nf
or
m
at
i
c
s
, vol
. 16, no. 4, pp. 21
89
–
2201, 2020, doi
:
10.1109/
T
I
I
.2019.2935153.
[
7]
L
.
F
u,
S
.
Z
ha
ng,
Y
.
G
ong,
a
nd
Q
.
H
ua
ng,
“
M
e
di
c
i
ne
gl
a
s
s
bot
t
l
e
de
f
e
c
t
de
t
e
c
t
i
on
ba
s
e
d
on
m
a
c
hi
ne
vi
s
i
on,”
i
n
2019
C
hi
ne
s
e
C
ont
r
ol
a
nd D
e
c
i
s
i
on
C
onf
e
r
e
n
c
e
(
C
C
D
C
)
, 2019, pp. 5681
–
5685, doi
:
10.1109/
C
C
D
C
.2019.8832688.
[
8]
R
.
K
ul
ka
r
ni
,
S
.
K
ul
ka
r
ni
,
S
.
D
a
bha
ne
,
N
.
L
e
l
e
,
a
nd
R
.
S
.
P
a
s
w
a
n,
“
A
n
a
ut
o
m
a
t
e
d
c
om
put
e
r
vi
s
i
on
-
ba
s
e
d
s
y
s
t
e
m
f
or
bot
t
l
e
c
a
p
f
i
t
t
i
ng
i
ns
pe
c
t
i
on,”
2019
T
w
e
l
f
t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
C
ont
e
m
por
ar
y
C
om
put
i
ng
(
I
C
3)
,
N
oi
da
,
I
ndi
a
,
2019,
pp.
1
-
5,
doi
:
10.1109/
I
C
3.2019.8844942.
[
9]
C
.
H
e
,
C
.
L
i
,
B
.
C
he
n,
B
.
Y
ua
n,
a
nd
Y
.
Y
i
n,
“
R
e
s
e
a
r
c
h
on
de
f
e
c
t
de
t
e
c
t
i
on
of
t
he
out
e
r
s
i
de
of
bot
t
l
e
c
a
p
ba
s
e
d
on
hi
gh
a
ngl
e
a
nd
m
ul
t
i
-
vi
e
w
vi
s
i
on s
ys
t
e
m
,”
I
E
E
E
A
c
c
e
s
s
, vol
. 11, pp. 65798
–
65809, 2023, doi
:
10.1109/
A
C
C
E
S
S
.2023.3290616.
[
10]
M
.
K
a
z
m
i
,
B
.
H
a
f
e
e
z
,
H
.
R
.
K
ha
n,
a
nd
S
.
A
.
Q
a
z
i
,
“
M
a
c
hi
ne
-
vi
s
i
on
-
ba
s
e
d
pl
a
s
t
i
c
bot
t
l
e
i
ns
p
e
c
t
i
on
f
or
qua
l
i
t
y
a
s
s
ur
a
nc
e
,”
i
n
E
ngi
ne
e
r
i
ng P
r
oc
e
e
di
ngs
, 2022, vol
. 20, no. 1, doi
:
10.3390/
e
ngpr
oc
20220200
09.
[
11]
C
.
L
i
,
Y
.
Y
i
n,
B
.
Y
u
a
n,
a
nd
X
.
L
i
,
“
R
e
s
e
a
r
c
h
on
s
ur
f
a
c
e
de
f
e
c
t
de
t
e
c
t
i
on
t
e
c
hnol
ogy
ba
s
e
d
on
t
he
z
on
a
l
a
nd
t
i
m
e
-
s
ha
r
i
n
g
c
om
put
a
t
i
ona
l
i
m
a
gi
ng,”
I
E
E
E
A
c
c
e
s
s
, vol
. 10, pp. 79574
–
79583, 2022, doi
:
10.1109/
A
C
C
E
S
S
.2022.3163726.
[
12]
S
.
B
.
J
ha
a
nd
R
.
F
.
B
a
bi
c
e
a
nu,
“
D
e
e
p
C
N
N
-
ba
s
e
d
vi
s
ua
l
de
f
e
c
t
de
t
e
c
t
i
on:
s
u
r
ve
y
of
c
ur
r
e
nt
l
i
t
e
r
a
t
ur
e
,”
C
om
put
e
r
s
i
n
I
ndus
t
r
y
,
vol
. 148, 2023, doi
:
10.1016/
j
.c
om
pi
nd.2023.103911.
[
13]
T
.
A
.
K
or
z
he
bi
n
a
nd
A
.
D
.
E
gor
ov,
“
C
om
pa
r
i
s
on
of
c
om
bi
na
t
i
ons
o
f
da
t
a
a
ugm
e
nt
a
t
i
on
m
e
t
hods
a
nd
t
r
a
ns
f
e
r
l
e
a
r
ni
ng
s
t
r
a
t
e
gi
e
s
i
n
i
m
a
ge
c
l
a
s
s
i
f
i
c
a
t
i
on
us
e
d
i
n
c
onvol
ut
i
on
de
e
p
ne
ur
a
l
ne
t
w
or
ks
,”
2021
I
E
E
E
C
onf
e
r
e
nc
e
of
R
us
s
i
an
Y
oung
R
e
s
e
a
r
c
he
r
s
i
n
E
l
e
c
t
r
i
c
al
and E
l
e
c
t
r
oni
c
E
ngi
ne
e
r
i
ng (
E
l
C
onR
us
)
, pp. 479
–
482, 2021, doi
:
10.1109/
E
l
C
onR
us
51938.2021.9396724.
[
14]
K
.
L
i
u,
Y
.
L
i
u,
C
.
P
e
ng,
Y
.
C
ha
ng,
a
nd
Y
.
Z
ha
o,
“
D
e
s
i
gn
of
ha
r
dw
a
r
e
a
c
c
e
l
e
r
a
t
i
on
i
n
e
dge
c
om
put
i
ng
de
vi
c
e
f
or
bot
t
l
e
c
a
p
hi
gh
-
s
pe
e
d i
ns
pe
c
t
i
on,”
W
i
r
e
l
e
s
s
C
om
m
uni
c
at
i
on
s
and M
obi
l
e
C
om
put
i
ng
, vol
. 2022, 2022, doi
:
10.1155/
2022/
5270887.
[
15]
H
.
X
u,
Y
.
X
u,
a
nd
K
.
H
u,
“
A
vi
s
i
on
-
ba
s
e
d
i
ns
pe
c
t
i
on
s
ys
t
e
m
f
or
pha
r
m
a
c
e
ut
i
c
a
l
pr
oduc
t
i
on
l
i
ne
,”
J
our
nal
of
Supe
r
c
om
put
i
ng
,
vol
. 81, no. 4, 2025, doi
:
10.1007/
s
11227
-
025
-
07135
-
8.
[
16]
H
.
L
v,
S
.
H
ua
ng,
W
.
X
i
e
,
Z
.
C
he
n,
Z
.
G
uo,
a
nd
X
.
Z
ha
ng,
“
M
ul
t
i
-
s
c
a
l
e
ba
s
e
d
de
f
e
c
t
de
t
e
c
t
i
on
f
or
a
ut
om
ot
i
ve
g
l
a
s
s
,”
J
our
nal
of
P
hy
s
i
c
s
:
C
onf
e
r
e
nc
e
Se
r
i
e
s
, vol
. 2303, no. 1, 2022, doi
:
10.1088/
1742
-
6596/
2303/
1/
012068.
[
17]
Y
.
Z
ha
ng,
S
.
S
hu,
X
.
L
a
ng,
H
.
L
i
a
ng,
Z
.
Y
u,
a
nd
Z
.
Y
a
ng,
“
A
r
e
a
l
-
t
i
m
e
m
e
t
h
od
f
or
de
t
e
c
t
i
ng
bot
t
om
de
f
e
c
t
s
of
l
i
t
hi
um
ba
t
t
e
r
i
e
s
ba
s
e
d
on
a
n
i
m
pr
ove
d
Y
O
L
O
v5
m
ode
l
,”
M
e
as
ur
e
m
e
nt
Sc
i
e
nc
e
and
T
e
c
hn
ol
ogy
,
vol
.
34,
no.
12,
2023,
do
i
:
10.1088/
1361
-
6501/
a
c
f
9bf
.
[
18]
X
.
C
he
n,
Z
.
J
i
a
ng,
H
.
C
he
ng,
H
.
Z
he
ng,
a
nd
Y
.
D
u,
“
L
F
-
Y
O
L
O
v4:
a
l
i
ght
w
e
i
ght
de
t
e
c
t
i
on
m
ode
l
f
o
r
e
nha
nc
i
ng
t
he
f
us
i
on
of
i
m
a
ge
f
e
a
t
ur
e
s
of
s
ur
f
a
c
e
de
f
e
c
t
s
i
n
l
i
t
hi
um
ba
t
t
e
r
i
e
s
,”
M
e
as
ur
e
m
e
nt
S
c
i
e
nc
e
and
T
e
c
hnol
ogy
,
vol
.
35,
no.
2,
2024,
doi
:
10.1088/
1361
-
6501/
a
d0690.
[
19]
J
.
L
i
u
a
nd
D
.
Z
ha
ng,
“
R
e
s
e
a
r
c
h
on
ve
hi
c
l
e
obj
e
c
t
de
t
e
c
t
i
on
a
l
gor
i
t
hm
ba
s
e
d
on
i
m
pr
ove
d
Y
O
L
O
v3
a
l
gor
i
t
hm
,”
i
n
J
our
nal
o
f
P
hy
s
i
c
s
:
C
onf
e
r
e
nc
e
Se
r
i
e
s
, 2020, vol
.
1575, no. 1, doi
:
10.1088/
1742
-
6596/
1575/
1/
012150.
[
20]
L
.
S
he
n,
H
.
T
a
o,
Y
.
N
i
,
Y
.
W
a
ng,
a
nd
V
.
S
t
oj
a
novi
c
,
“
I
m
pr
ove
d
Y
O
L
O
v3
m
o
de
l
w
i
t
h
f
e
a
t
ur
e
m
a
p
c
r
oppi
ng
f
or
m
ul
t
i
-
s
c
a
l
e
r
oa
d
obj
e
c
t
de
t
e
c
t
i
on,”
M
e
as
ur
e
m
e
nt
Sc
i
e
nc
e
and T
e
c
hnol
ogy
, vol
. 34, no. 4, 2023,
doi
:
10.1088/
1361
-
6501/
a
c
b075.
[
21]
Z
.
H
ua
ng,
H
.
H
u,
Z
.
S
he
n,
Y
.
Z
ha
ng,
a
nd
X
.
Z
ha
ng,
“
L
i
ght
w
e
i
ght
e
dg
e
-
a
t
t
e
nt
i
on
ne
t
w
or
k
f
or
s
ur
f
a
c
e
-
de
f
e
c
t
de
t
e
c
t
i
on
of
r
ubbe
r
s
e
a
l
r
i
ngs
,”
M
e
as
u
r
e
m
e
nt
Sc
i
e
n
c
e
and T
e
c
hnol
ogy
, vol
. 33, no. 8, 2022, doi
:
10
.1088/
1361
-
6501/
a
c
6663.
[
22]
H
.
W
a
ng,
X
.
X
u,
Y
.
L
i
u,
D
.
L
u,
B
.
L
i
a
ng,
a
nd
Y
.
T
a
ng,
“
R
e
a
l
-
t
i
m
e
de
f
e
c
t
d
e
t
e
c
t
i
on
f
or
m
e
t
a
l
c
om
pone
nt
s
:
a
f
us
i
on
of
e
nh
a
nc
e
d
c
a
nny
–
de
ve
r
na
y a
nd Y
O
L
O
v6 a
l
gor
i
t
hm
s
,”
A
ppl
i
e
d Sc
i
e
nc
e
s
, vol
. 13, no. 12, 2023, doi
:
10.3390/
a
pp13126898.
[
23]
M
.
S
t
a
s
e
vyc
h
a
nd
V
.
Z
va
r
yc
h,
“
I
nnova
t
i
ve
r
obot
i
c
t
e
c
hnol
ogi
e
s
a
nd
a
r
t
i
f
i
c
i
a
l
i
nt
e
l
l
i
ge
nc
e
i
n
pha
r
m
a
c
y
a
nd
m
e
di
c
i
ne
:
pa
vi
ng
t
h
e
w
a
y
f
or
t
he
f
ut
ur
e
of
he
a
l
t
h
c
a
r
e
-
a
r
e
vi
e
w
,”
B
i
g
D
at
a
and
C
ogni
t
i
v
e
C
om
put
i
ng
,
vol
.
7,
no.
3,
A
ug.
2023,
doi
:
10.3390/
bdc
c
7030147.
Evaluation Warning : The document was created with Spire.PDF for Python.