TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.5, May 2014, pp
. 3737 ~ 37
4
4
DOI: http://dx.doi.org/10.11591/telkomni
ka.v12i5.5108
3737
Re
cei
v
ed
No
vem
ber 1
1
, 2013; Re
vi
sed
De
cem
ber 2
3
,
2013; Accep
t
ed Jan
uary 8
,
2014
Development of a Machine Vision System for Solar
Wafer Counting
Nan
Wang, Shuguang Z
h
a
ng, Man Ch
eng, Zhenjia
ng Cai*
Coll
eg
e of Elec
trical an
d Mech
anic
a
l Eng
i
n
e
e
r
ing, Agric
u
ltur
al Univ
ersit
y
of
Hebe
i,
Baod
ing 071
00
1,
Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: czj65@
163.c
o
m
A
b
st
r
a
ct
The trad
ition
a
l
man
ual
cou
n
t
ing w
a
fers
le
ade
d
to t
he
silico
n
w
a
fer
cracked
by
o
perati
n
g
freque
ntly. Instead
of the
manu
al w
o
rk, th
is pa
per
pro
p
o
sed
a syste
m
to
cou
n
tin
g
w
a
fers bas
ed
on
Machi
ne Visi
o
n
theory a
nd Ima
ge Proc
essi
ng al
gorit
hm.
We desig
ne
d a counter syst
em
and a
d
o
p
te
d
infrare
d
le
d as
para
lle
l il
lu
mi
natio
n so
urce.
In i
m
ag
e pr
e-
process
i
ng, th
i
s
pap
er pr
ese
n
ted a
seri
es
of
alg
o
rith
ms, w
h
ich contai
ned i
m
a
ge s
m
o
o
thi
ng, unev
en
i
m
age corr
ection
and i
m
a
ge
mor
pho
logy o
per
ation
.
T
h
is pap
er pro
pose
d
a vertic
al proj
ectio
n
counti
ng b
a
sed
on statistics analysis su
bstit
u
te for the Ho
ugh
straight li
nes d
e
tection, a
nd the metho
d
s ha
ve ac
hi
eve
d
id
eal effects by e
x
peri
m
e
n
tal re
sults.
Key
w
ords
:
u
neve
n
Illu
mi
nat
ion, i
m
a
ge pre-
process
i
ng, co
untin
g al
gorith
m
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Photoele
c
tri
c
conve
r
si
on e
fficiency, pro
duct
q
uality rate and
pro
d
uction
of co
st
are the
three
key fact
ors
whi
c
h pl
a
y
an importa
nt role in the
sola
r cell pro
ductio
n
pro
c
e
ss. Not only the
quality of ma
terial itself, but also
the
crack
ed wafe
r redu
ce
s
th
e quality
rate, and
the
pri
n
cipal
element i
m
p
a
cts the
cracked
wafer i
s
the tra
d
it
ion
a
l man
ual
co
unting
wafe
rs. Firstly, all
the
cou
n
ting
re
sults a
r
e
freq
uently ina
ccura
cy. Seco
ndly, due
to
the fragile
prop
ertie
s
of the
material, h
a
n
d
ope
ration
may lead to t
he dam
age
wafers
and th
e
eco
nomi
c
lo
sses. Fin
a
lly, for a
long time
a
n
d
bo
ring
ma
nual
co
untin
g may
re
sult
in fatigu
e o
perato
r
and
redu
ce
the
work
efficien
cy. To
solve
this a
bove d
r
a
w
ba
ck,
we
ad
opt
ed ma
chi
ne
vision m
e
tho
d
in
stead
of
the
manual
work,
which
can p
r
ovide a
c
cura
cy and efficie
n
cy in wafe
r cou
n
ting. In this arti
cle, the
machi
n
e
visi
on
frame
w
o
r
k
in
clu
d
e
s
speci
a
lly-de
s
ig
ned ma
chine
,
special p
a
rallel illuminat
ion
sou
r
ce, high
-resolution
ca
mera, p
e
rso
nal compute
r
and ima
ge
pro
c
e
ssi
ng t
e
ch
nolo
g
y. The
frame
w
ork
can improves countin
g th
e numb
e
rs
o
f
wafer qui
ckly, repeatedl
y, and accurately
while mini
mizing han
dling t
o
prevent wafer dam
age.
There a
r
e
so
me research
es
on
pap
er
cou
n
ting
usi
n
g ma
chin
e vi
sion
technol
o
g
y [1-3],
but it is very limited on the
study of sola
r wafe
r counting.
Som
e
pape
rs
wo
rke
d
on pa
p
e
r
cou
n
ting ba
sed on T
e
xture Featu
r
e [1-2], [1] pr
ese
n
ts a meth
od
based on
a
nalyzin
g texture
feature of th
e pap
er ima
g
e
, obtaining t
he bina
ry im
age throug
h
LOG filter. T
hen the
cou
n
t
ing
numbe
rs
we
re obtai
ned
by
pixel p
r
oj
ecti
on al
gorith
m
based
on tilt
corre
c
ting
of t
he im
age
an
d
by
differen
c
e al
gorithm b
a
se
d on statisti
c analysis
re
spectively. A
method resp
ectively base
d
on
2D Ga
bo
r filter an
d 1D lin
e-by-li
ne freq
uen
cy anal
ysi
s
is p
r
op
osed
in [2], then counting n
u
mb
ers
by extractin
g
the filtere
d
bord
e
r i
n
form
ation of
p
a
p
e
r. But when
the si
de
of pape
r h
a
s
b
een
abra
ded
which lea
d
to u
n
e
v
en in
sh
ado
w, the
re
sults in [1]
were e
rro
r frequ
entl
y
, while [2]
can
not satisfying
the real-tim
e
req
u
ire
m
ent
s .To
so
lve
th
ese
problem,
[3] pre
s
e
n
ts
a metho
d
b
a
s
ed
on mathem
atical mo
rphol
o
g
y, the result
s ca
n be
sati
sfied in pract
i
cal produ
ctio
n. A countin
g
method ba
se
d on textural
prope
rty [4], which ac
co
rding to the common featu
r
e of piece a
nd
wafer,
which
firstly lo
cat
ed
wafe
r a
r
ea e
m
ployin
g first-o
r
de
r statisti
cs m
e
thod
and
e
dge
detectio
n
p
r
o
j
ection, a
nd t
hen o
b
taine
d
the num
be
rs by extre
m
e value a
nal
ysis in
practi
cal
appli
c
ation.
The
stru
cture
of this pap
e
r
in
clud
ed: th
e
solar wafer co
unter sy
stem an
d the
parall
e
l
illumination
source
whi
c
h
rest
rain
ed th
e uneven ill
u
m
ination in fi
rst pa
rt. The
digital imag
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3737 – 37
44
3738
pro
c
e
ssi
ng m
e
thod an
d the cou
n
ting al
gorithm a
r
e
appli
c
ation fo
r sola
r wafe
r in second p
a
rt.
And the expe
riment re
sult
s and analy
s
is
are sho
w
n in
last part.
2. Solar Wafer Coun
ter S
y
stem
The sola
r
wafer counte
r
system
i
n
clu
des pr
ote
c
tive exterio
r
sy
stem, sl
eeve
sup
port
sy
st
em,
dis
p
l
a
y
sy
st
e
m
a
nd cou
n
t
i
ng sy
st
em.
The
rol
e
of th
e
prote
c
tive ex
terior sy
stem
is
prote
c
ting th
e came
ra from envi
r
onm
ent interf
e
r
e
n
ce
an
d hol
ding it in
th
e prope
r
po
sitio
n
steadily. T
he
feature
of the
sle
e
ve
sup
p
o
rt system
is assistin
g
o
p
e
r
ator in
loadi
n
g
an
d
unloa
di
ng
the wafe
r e
a
s
ily, and
prov
iding id
entica
l
sleeve
po
sit
i
oning. T
he d
i
splay
system
is
sho
w
in
g the
real
-time ima
ge and the computed result. The above
three syste
m
is the external sy
stem, and
the last co
un
ting system i
s
the intern
a
l
sy
stem, wh
ich rol
e
is i
m
age p
r
o
c
e
s
sing a
nd ima
g
e
analyzi
ng. Th
e sch
e
matic
diagram
of th
e sy
stem i
s
shown in
Figu
re 1(a),
and
th
e a
c
tual
pro
d
u
ct
is sh
own in Figure 1
(
b
)
.
Figure 1(a
)
. T
he Sche
matic Diagram of Cou
n
ter Syst
em
(1)
(2)
Figure 1(b
)
: (1) The Protective Exterior Syst
em and
Sleeve Supp
ort System, (2) The A
c
tual
Produ
ct
Acco
rdi
ng to
the experi
m
e
n
t, the came
ra is
imp
a
cte
d
by the illumi
nant in the p
r
otective
exterior
syste
m
. Whe
n
un
e
v
en lumina
nce, the wa
fer
bord
e
r
ca
n n
o
t be
captu
r
e
d
by the
cam
e
ra
due to
the l
o
w
contrast
an
d lo
w fre
que
n
c
y noi
se
s.
In
the present
study,
a sealin
g da
rk box
was
desi
gne
d to avoid this prob
lem. The cam
e
ra is fi
xed in
the back of the box
inner,
and the front is
cutting
a
squ
a
re
gap
to th
e came
ra
ca
pture
imag
e.
In the
right
o
f
the b
o
x inn
e
r, the
pa
rall
el
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
De
velo
pm
ent of a Machine
Vision Syste
m
for Solar Wafer Co
unting
(Nan
Wan
g
)
3739
illumination source is placed to pr
ovide even illumination.
To
avoi
d the noise affected by visible
spectrum, we adopt infr
ared led i
n
stead of common l
ed. Th
e illumi
nation
schem
atic di
agram i
s
sho
w
n in Fig
u
re 2
(
a), an
d the dark box in Figure 2(b
)
.
Figure 2(a
)
. T
he Parall
el Illumination Pri
n
cipl
e
Figure 2(b
)
. T
he Sealing
Dark Box
3. Image Processing
for
Solar Wafer
The me
cha
n
i
s
m of employ
image proce
ssi
ng to co
un
t the wafers n
u
mbe
r
wa
s ill
ustrate
d
in the following c
h
art (Figure 3).
Figure 3. Image Pro
c
e
ssi
n
g
Flowcha
r
t
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3737 – 37
44
3740
After the
cam
e
ra
ca
pture
a
n
imag
e of
solar
wafe
r, th
e ori
g
inal i
m
a
ge i
s
RGB im
age,
so
the first p
r
ocedure is
ne
cessary to tra
n
sform the
o
r
iginal im
age
to gray ima
ge (Fi
gure 4(a)),
whi
c
h also re
mains mo
st o
f
the useful inform
ation an
d do not affect the sub
s
eq
uent pro
c
e
s
si
ng
[5].
3.1. Image Smoothing
Since th
e noi
se exi
s
ts i
n
surroun
ding
a
nd sola
r wa
fer, it is essenti
a
l
for
restraining it in
the pro
c
e
s
s
of image p
r
e
p
ro
ce
ssi
ng.
Median
Filter
[6] is used i
n
[4], which
can
eliminate
the
isolate noi
se effectively,
but
it
not use for the
strip
e
analysi
s
. A
ccordin
g to the
hori
z
ontal
stri
pe
feature
of th
e obj
ect im
a
ge, Avera
ge
Filter [6]
i
s
u
s
ed
to
smoot
h the
pictu
r
e
,
instea
d of t
h
e
conve
n
tional 3×3
squa
re kernel, 1×3 ke
rnel
to
a
c
com
m
odate th
e i
m
age
wa
s
be
ado
pted,
whi
c
h
can be elimi
n
ating pixel values which are unre
p
re
se
n
t
ative of
their surro
undi
ngs.
Meanwhile the
filtering al
go
rithm ca
uses
image fu
zzy, so
t
he i
m
proved Sob
e
l [
6
] ope
rato
r i
s
u
s
e
d
in
ed
ge
detectio
n
in o
r
de
r to sha
r
p
en the image.
The Sobel
op
erato
r
is a di
screte
different
iation ope
rato
r, comp
uting
an app
roxima
tion of
the gra
d
ient
of the image
intensity fun
c
tion. Th
e Sobel o
perator is ba
se
d on
convolvin
g
th
e
image with
a small, sepa
ra
ble, and integ
e
r valued f
ilte
r
in hori
z
o
n
tal
and vertical dire
ction an
d is
therefo
r
e
rela
tively inexpen
sive in te
rm
s
of com
putatio
ns. Be
cau
s
e
the ho
rizontal
feature
of th
e
sola
r wafer,
we u
s
e th
e i
m
prove
d
Sob
e
l hori
z
o
n
tal
operator (a
s 2-2
)
to
in
st
ea
d of the defa
u
lt (as
2-1
)
, which i
s
fitting fo
r t
he im
age
preferably. T
h
e
re
sult i
s
sh
o
w
n i
n
Fi
gure
2-3, comp
arison
Figure 4
(
b
)
a
nd Fi
gure
4(c), the
improved m
e
thod
ca
n en
han
ce
th
e lo
w
co
ntra
st part
is supe
rior
to the default operator.
Default Sobel
horizontal op
erato
r
:
12
1
00
0
12
1
(1)
Improved So
bel hori
z
o
n
tal
operato
r
:
11
2
1
1
11
2
1
1
0
000
0
11
2
1
1
11
2
1
1
(
2
)
Figure 4(a
)
. T
he Origi
nal G
r
ay
Image
Figure 4(b
)
. Default Sobel
Operator
Figure 4(c). Improve
d
Sob
e
l
Operator
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
De
velo
pm
ent of a Machine
Vision Syste
m
for Solar Wafer Co
unting
(Nan
Wan
g
)
3741
3.2. Unev
en Image Corr
ec
tion
Ahead of ima
ge co
rrectio
n
, it must extract t
he wafer f
r
om ba
ckgro
und. As the theory of
mean value
of the image
can m
easure
luminan
ce, fi
rstly, taking h
o
rizontal p
r
oj
ection by me
an
value of
ea
ch
ro
w,
we
den
ote by m
a
trix
()
M
i
. Secondly,
we ave
r
agi
ng
()
M
i
and th
e
re
sul
t
is A.
Last, ope
rati
ng formul
a
()
Ni
=
()
M
i
-A, and
()
Ni
is a nonn
egati
v
e numbe
r. Figure 5(a)
can
illustrate the procedure,
()
M
i
is sh
own in the upp
er g
r
a
ph, and
()
Ni
is di
splaye
d in th
e unde
r
grap
h. We e
x
tract wafer f
r
om b
a
ckg
r
o
und u
s
in
g e
d
ge info
rmatio
n of the im
a
ge fro
m
ab
o
v
e
method. The
result of imag
e extraction i
s
sh
own in Figure 6.
Figure 5. Image Lo
cation
Method
In spite of
u
s
ing th
e infrared
pa
rallel
illumination
to solve the
influen
ce of
uneve
n
luminance for the solar wa
f
e
r, there i
s
st
ill some
low contrast part i
n
picture, as
shown in green
squ
a
re
of Fi
g
u
re
6(a).
Re
move the
bia
s
lig
ht from
t
he o
r
igin
al im
age i
s
an
effective m
e
tho
d
to
uneven illumi
nation corre
c
tion, su
ch a
s
estimating b
a
c
kgro
und
with morp
holo
g
i
c
al op
eratio
n [6],
and usi
ng a
polynomial to fitting bia
s
light. We adopt sp
atial
filtering to
corre
c
t unev
en
illumination, i
n
ad
dition, thi
c
kne
s
s of o
n
e
pie
c
e
is
ab
out 5
~
8
pixels, so u
s
in
g th
e imp
r
oved
Sobel
operation (fo
r
mula 2-2
)
to filtering the ob
ject
image, which a
c
hieve
d
a good re
sult as displ
a
yed
in gree
n sq
ua
re of Figu
re 6
(
b).
Figure 6(a
)
. L
o
w Contra
st (gree
n
sq
ua
re
)
Fi
gure 6(b
)
. Uneve
n
Co
rrection (green
squ
a
re
)
3.3. Image Morpholog
y
Sobel o
p
e
r
at
or i
s
a
di
screte diffe
rent
iation o
perator,
so the
pro
c
e
s
sed i
m
age i
s
discon
ne
cted
after filterin
g. To
solve
t
he problem,
the op
enin
g
and
clo
s
in
g
ope
ration f
r
om
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3737 – 37
44
3742
mathemati
c
al
morp
holo
g
y are u
s
e
d
. Th
e ope
ning
op
eration i
s
d
e
fined a
s
e
r
o
s
i
on follo
wed
by
dilation u
s
ing
the same
structuri
ng ele
m
ent for both
o
peratio
ns. O
n
the contrary, it is the closi
ng
operation. A
c
cording
to th
e expe
rime
nt analy
s
is
an
d
the
hori
z
o
n
tal cha
r
a
c
ter
of the im
age,
we
adopt 1
×
3, 1
×
7 a
nd 1
×
1
1
stru
cturi
ng el
ement is
use
d
for this
step
, and the 1
×
7
element
can
be
suit for morp
hology ope
rat
o
r (Fig
ure 7
)
.
The di
scon
n
e
cted a
r
ea of
the image was shown in re
d
roun
d of Figu
re 7(a), in ad
dition the re
sult
of image
morp
holo
g
y operatin
g was
displ
a
yed in red
roun
d of Figu
re 7(b).
Figure 7(a
)
. Disco
nne
cted
Image (red round
)
Figure 7(b
)
. Image Mo
rph
o
logy (red ro
und)
3.4 Coun
ting
Algorithm for Solar Wafer
Hou
gh tran
sf
orm [7] is the comm
on m
e
thod for det
ecting
straig
h
t
lines, in the image
spa
c
e, the st
raight line can
be describe
d
as
ya
x
b
where the parameter ‘
a
’ is the slo
p
e
of
the line and
‘b’ is the inte
rce
p
t. This i
s
called the
sl
ope-i
n
terce
p
t model of a
straig
ht line. In
gene
ral, the
Hou
gh tra
n
sf
orm
s
ch
ang
e
s
into the
pol
ar coordinate
spa
c
e,
an
d the equ
ation o
f
the
line is written as:
co
s
s
i
n
rx
y
The pa
ramet
e
r
r
repre
s
e
n
ts the dista
n
ce betwe
en th
e line and the
origin, while
is t
h
e
angle of the
vector from
the origin t
o
this clo
s
e
s
t point. The plane of (
r
,
) is sometimes
referred to a
s
Houg
h sp
ace for the set o
f
straight line
s
in two dime
nsio
ns.
The Houg
h d
e
tection
nee
d
s
trave
r
si
ng e
a
ch
pixel and
confirming th
e pro
p
e
r
(
r
,
), s
o
it
is p
o
o
r
p
e
rfo
r
mance i
n
effi
cien
cy a
n
d
a
daptabilit
y fo
r co
unting
n
u
m
bers
(results in
Ta
ble
1).
So
the pape
r pro
posed a ne
w
method which bas
ed on
statistics analy
s
is illu
strate
d:
Step 1: Th
e
mono
ch
rome
image
after
pro
c
e
ssi
ng
b
y
step
2-3
re
gard
s
as M,
casting
out
both the
left
and
right
20
colum
n
s pixel
s
, which a
r
e
more
apt to
d
i
sturbi
ng
by the n
o
ise tha
n
the
centre pixels,
and the re
sul
t
as N.
Step 2: Dividing the image
N into 5 su
b-i
m
age
s equ
all
y
, which re
ga
rd as
N1
~N5.
Step 3: Acco
rding to
the h
o
rizontal feat
ure of
th
e wa
fer, usi
ng the
vertical
proj
e
c
tion to
each column
of N
i
(
i
=1
~5
), when
cu
rre
nt pi
xel is 1, a
nd t
he
next
2 pixel
is
0, so re
g
a
rd as
th
ere
is a line in current po
sition ,the lines n
u
m
ber ad
d 1,obt
aining the ma
trix R
i
(
i
=1~
5
).
Step 4: Usin
g
the maj
o
rity
voting by stat
istics a
nalysi
s
, finding the
most time
s n
u
mbe
r
in
R
i
(
i
=1
~5
) and
storin
g the
numbe
r in r
i
(
i
=1
~5). Simila
rly
for the r
i
(
i
=1~5
) processing t
h
e
majority voting and getting
the result C, whi
c
h is the h
o
rizontal line
s
’
The re
sult
s with above method are sh
own in Table 2.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
De
velo
pm
ent of a Machine
Vision Syste
m
for Solar Wafer Co
unting
(Nan
Wan
g
)
3743
4. Results a
nd Analy
s
is
In Table 1, th
e (30
50 80
100 12
0 13
5
150)
sam
p
le
s were sepa
ra
tely counted i
n
non
-
repe
at 10 tim
e
s b
a
sed on t
he metho
d
of
the pape
r p
r
opo
sed. Th
e
sampl
e
corre
c
t co
unting
ra
tio
is ab
out 9
8
%
whe
r
e
10
,
1
0
&
2
r
, but it spen
ds
over
3 seco
nd
whe
n
the sampl
e
numbe
rs i
s
1
50. By co
ntra
st Ta
ble 2, th
e (5
0
1
00 15
0) sam
p
le
s were se
pa
ratel
y
cou
n
ted
i
n
non-
repe
at 5 times based on th
e theory of Hough dete
c
tio
n
straig
ht lines. The co
rrect ratio is nearl
y
100% and h
a
s
a goo
d agin
g
. In addition, the 50, 100 and 15
0 sam
p
les
were se
parately coun
ted
100 pa
ckage
s witho
u
t rep
eating, and t
he re
sults a
s
sho
w
n in Fi
gure 8
(
a
)
to Figure 8(c). It
achi
eved id
e
a
l effect
s eq
u
i
valently and
the co
rrec
t
ra
tio is ove
r
9
5
%
, so the
co
unting al
go
rithm
based on
statistics analy
s
is was a
dapte
d
to solar
wafe
r cou
n
ting.
Table 1. Ho
u
gh Co
unt Nu
mbers
10
,
1
0
&
2
r
Sample
Count
Hough Co
unt Wa
fer Numbe
r
s Non
-
Repeat
10 Time
s
Mean
Ti
me
30
30 30
30 30
30 30
30 30
30 30
1s
50
50 50
50 50
50 50
50 50
50 50
1.2s
80
80 80
80 80
80 80
79 80
80 80
1.4s
100
100 100
100
99
100 100
100 100
101 100
1.6s
120
120 120
120 119
120 121
120 120
120 120
1.9s
135
135 135
134 135
133 135
135 135
136 135
2.4s
150
150 151
150 150
150 149
150 150
150 150
3.1s
Table 2. Parti
t
ioning Verti
c
al Obje
ction
Cou
n
t Numb
ers
Sample
Count
Non-Re
peat
5 Times
r1
r2
r3
r4
r5
C
Ti
me
50
50-1
50 50
49 50 50 50
1.2s
50-2
50 51
50 50 50 50
1.2s
50-3
49 50
50 50 49 50
1.2s
50-4
50 49
50 50 50 50
1.2s
50-5
49 50
50 50 51 50
1.2s
100
100-1
101 100
100 101 100 100
1.3s
100-2
100
100
100 99
99 100
1.3s
100-3
99
99
100 98 100
100
1.3s
100-4
100 101
100 100 101 100
1.3s
100-5
100 100
101 100 101 100
1.3s
150
150
150-1
149 150
150 150 150 150
1.5s
150-2
149 151
152 150 150 150
1.5s
150-3
150 151
150 149 150 150
1.5s
150-4
150 150
150 148 151 150
1.5s
150-5
151 150
150 152 149 150
1.5s
Figure 8(a
)
. 5
0
Piece
s
Figure 8(b
)
. 1
00 Piece
s
Figure 8(c). 1
50 Piece
s
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 5, May 2014: 3737 – 37
44
3744
5. Conclusio
n
This
pap
er
p
r
esented
ima
ge p
r
o
c
e
ssi
n
g
alg
o
rithm
s
:
image
sm
oo
thing, un
eve
n
imag
e
corre
c
tion
an
d image m
o
rpholo
g
y ope
ration. This
pa
per p
r
o
p
o
s
ed
a vertical
projectio
n
coun
ting
based on
sta
t
istics a
nalysi
s
su
bstitute f
o
r the H
oug
h
straig
ht line
s
dete
c
tion.
By experime
n
ta
l
results, the m
e
thod
s have
achi
eved ide
a
l effects an
d
corre
c
t ratio.
Referen
ces
[1]
MIAO Lian
g,
PING Xi
ji
an
.
Algorit
hm of
Paper
Co
unti
n
g
Bas
ed on
T
e
xture
F
eatu
r
e.
Jour
nal
of
Information En
gin
eeri
ng U
n
iv
ersity
. 2005; 6(
4): 47-50.
[2]
LI Yi, RUAN
Qiuqi. Al
gorith
m
of P
aper C
ounti
ng B
a
sed
on T
e
xture A
nal
ysis.
J
ourn
a
l
of Ima
ge
an
d
Graphics
. 20
04
; 9(9): 1042-1
0
48.
[3]
Z
H
EN Guan
g, CHEN Ya
op
i
ng, YU W
e
n
y
ong. T
he
stud
y of a
l
g
o
rithm
of pap
er cou
n
ting
base
d
o
n
mathematic
al morph
o
lo
g
y
.
Mi
cro
c
om
pu
te
r
Information.
20
07; 23(2
1
): 214
-216.
[4] FANG
Chao
，
TA
N
W
e
i
，
DU Jian-
ho
ng. S
o
lar
en
erg
y
w
a
fer cou
n
tin
g
base
d
o
n
te
xt
ural
prop
ert
y
.
Information a
n
d
Electron
ic En
gin
eeri
ng.
20
1
1
; 9(2); 185-1
8
9
.
[5]
Jean-
Luc Star
ck, F
i
onn M
u
rtagh, Em
m
anu
el
J Can
dès, David
L Do
no
ho.
Gray
an
d
C
o
l
o
r Ima
g
e
Contrast En
ha
nceme
n
t b
y
t
h
e Curve
l
et T
r
ansform.
IEEE Transaction on Im
age Processing
. 2
003;
12(6): 70
6- 71
7.
[6]
Rafael C, Richard E.
Di
gita
l Ima
ge Pr
ocess
i
ng.
2
nd
ed. T
r
anslate
d b
y
RU
AN Qiuq
i. Bei
j
i
ng: Pu
blis
hin
g
Hous
e of Electronics Ind
u
str
y
.
2006.
[7] https://en.
w
i
k
i
p
edi
a.org/
w
i
k
i
/H
oug
h_tra
nsfor
m
[8] https://
w
w
w
.
ma
th
.
w
a
s
hi
ng
to
n.e
d
u
[9]
Yuel
in Z
ou,
Xi
aoq
ian
g
Li
an
g,
T
ongmin
g W
ang. Vi
sib
l
e a
n
d
Infrared Ima
ge F
u
sio
n
Usi
ng the L
i
ftin
g
Wavelet.
T
E
LKOMNIKA Indon
esia
n Journ
a
l o
f
Electrical Eng
i
ne
erin
g
. 201
3; 11(11): 62
90-
629
5.
[10] Yushu
Xio
ng.
A Cluster
ing
A
l
gorit
hm Base
d on
Ro
ugh
S
e
t and
Gen
e
ti
c Algor
ithm
.
TELKOMNIKA
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2013; 1
1
(10):
578
2-57
88.
Evaluation Warning : The document was created with Spire.PDF for Python.