TELKOM
NIKA
, Vol. 11, No. 4, April 2013, pp. 2124
~21
3
0
ISSN: 2302-4
046
2124
Re
cei
v
ed
Jan
uary 15, 201
3
;
Revi
sed Fe
br
ua
ry 26, 20
13; Accepted
March 6, 201
3
Printing Detecting Algorithm Based on
Maximum
Degree of Recognition
Hu Zhan
g
*
1
, Feng Guo
2
1,2
Key
Lab. of Broad
ba
nd
w
i
r
e
less comm
uni
cations
a
nd Se
nsor net
w
o
rks, W
uhan U
n
iver
sit
y
of
T
e
chnolog
y, W
uha
n, Hub
e
i, C
h
in
a
*Corres
p
o
ndi
n
g
author e-m
a
il
: xxxyk
y
b
@
w
h
ut.edu.cn, gu
ofeng.
w
h
ut@gm
a
il.com
A
b
st
r
a
ct
In mod
e
rn
pac
kagi
ng, pr
intin
g
in
dustry, d
u
e
to effe
cts of t
he pr
op
erties
of
the strip
itse
lf and
th
e
ambi
ent li
ght, strip back
g
rou
nd co
lor a
nd t
he co
lor of
the
printin
g
li
ne, t
he l
o
w
contras
t
boun
dari
e
s of
the
strip on b
o
th
sides a
nd s
o
on, the trad
itio
nal d
i
g
i
ta
l q
ual
itative d
e
tectio
n an
d
contro
l
to the correcti
o
n
system d
oes
not meet the
compre
hens
ive
requir
e
me
nts. This pa
per a
i
ms
to study t
he d
e
tectio
n o
f
a
contin
uo
us li
n
e
, disco
ntin
uo
us li
ne
and
c
o
lor
divi
di
n
g
l
i
ne o
n
the
stri
p, and
bec
aus
e of l
o
w
contrast
betw
een b
a
ckg
roun
d col
o
r an
d divi
din
g
li
ne,
w
e
propos
ed
a
n
inn
o
vativ
e
so
lutio
n
an
d i
m
pl
ementati
on. T
h
is
article
discuss
es a n
e
w
alg
o
r
ithm
bas
ing
o
n
maxi
mu
m d
egre
e
of rec
o
gniti
on
and
op
tima
l li
ght so
u
r
ce
search
al
gorith
m
, a
nd w
e
si
mu
late
d this
i
n
MAT
L
AB,
fi
nally, w
e
c
o
mplete
d the
phy
sical testi
ng
of th
e
overall system
.
Ke
y
w
ords
: printing, correction system
, degr
ee of
recognition, lig
ht source search
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
With
the co
ntinuou
s
d
e
vel
opment and appli
c
ati
on
of sci
entific a
n
d
tech
nolo
g
ical level,
the deg
ree
of automation
of indu
strial p
i
peli
ne
stre
ng
thened. In m
odern pa
ckag
ing and
pri
n
ting
indu
stry, flexible packa
gin
g
(su
c
h a
s
pl
astic p
r
in
ting,
slitting mach
ine, coating
machi
ne, prin
ting
and
dyeing
)
basi
c
pro
d
u
c
tion line
nea
rly finish
aut
omation.
Ho
wever,
unev
en thi
c
kne
s
s and
tensio
n incon
s
iste
nt made
quality deviation and
run
n
in
g deviation, therefo
r
e
strip
appea
rs lateral
deviation, dislocatio
n and
can
not ke
ep
a strai
ght lin
e
runni
ng. The
r
efore, in ord
e
r to solve th
is
seri
ou
s pro
b
l
e
m affecting the level of automati
on of modern pa
ckag
ing and
pri
n
ting indu
stry, the
need fo
r a
corrective
cont
rol d
e
vice
onl
ine tra
c
king
and p
r
om
ptly co
rre
cted
th
e po
sition of
the
hori
z
ontal di
rection of the
strip du
rin
g
transmi
ssion[1]
.
Corre
c
tive control
syste
m
co
nsi
s
ts
of
three
co
mpone
nts: t
he corre
c
tive po
sition
detecto
r, the
co
rre
ctive
controlle
r a
n
d
driver. T
he
c
o
rrec
tive pos
i
tion dete
c
t
or
in
th
e fo
r
e
mo
s
t
end of the co
rre
ction
control system
, an
d its main fun
c
tion is to det
ect the actual
position of the
strip m
a
terial
needi
ng
corre
c
tion in th
e h
o
rizontal di
re
ction. The
ma
in functio
n
of
the co
ntrolle
r
is
to re
ceipt
sig
nal from
corrector, to
dete
r
mine th
e
stri
p po
sition
an
d offset to
ref
e
ren
c
e
po
siti
on of
the di
re
ction,
and
calculat
e the
offset
signal. T
he m
a
in fun
c
tion
o
f
the exe
c
utio
n me
ch
anism
is
the reception
controll
er of the
con
d
itioni
ng sig
nal, co
mpletes t
he
correct process for gen
erating
offset strip. T
he main fun
c
tion of the execut
ion me
chani
sm is th
e re
ception t
he co
nditioni
ng
sign
al of cont
rolle
r, compl
e
tes the co
rrect proce
s
s [2-3
] .
This p
ape
r i
m
prove
d
det
ecting m
e
tho
d
of existing
corre
c
tive po
sition, and
we pro
p
o
s
ed
printing
dete
c
ting an
d
co
rrect p
o
sitio
n
a
l
gorithm
ba
si
ng o
n
m
a
ximum d
egree
of re
co
gnition
a
n
d
optimal light.
2. Hard
w
a
re
Implementation
The corre
c
tive positio
n d
e
tection
syst
em
diag
ram
is sh
own in Figure 1, du
ring the
corre
c
tive image acqui
sition pro
c
e
s
s, the STM32F
1
03RCT6 mo
d
u
le as the co
re functio
n
which
is to co
ntrol t
he CM
OS im
age sen
s
o
r
a
nd TFT L
C
D displ
a
y modu
le. The CM
O
S
image sen
s
or
module via a
bi-di
r
e
c
tional
transceive
r
di
splay
s
an im
age colle
ctin
g by CMOS on the scree
n
. In
the p
r
o
c
e
ss,
TFT L
C
D
can
display the
li
ne o
r
edge
n
eed
s to
be i
d
entified. STM
32
read
s ima
ge
data and
con
v
erts data int
o
the format of RGB565.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Printing Dete
cting Algo
rith
m
basing on
Maxim
u
m
Degree of Reco
gnition (Hu Z
hang
)
2125
Figure 1. Block
Diag
ram o
f
Correc
tive P
o
sition
Dete
ction System
Duri
ng the
proce
s
s when
corre
c
tive im
age
contra
st is
in
cre
a
se, STM32 will
a
nalysi
s
of
the colo
r co
ntrast, an
d co
ntrol the lumin
o
u
s inten
s
it
y of the light sou
r
ce m
odul
e. Furthe
r, we
will
find the
maxi
mum lu
mino
us i
n
ten
s
ity o
f
the
co
lo
r
contra
st by
a
binary
search meth
od. Im
age
recognitio
n
p
r
ocess,
an i
m
porta
nt task of
STM32
i
s
to
ide
n
tify colle
cted
ima
ge d
a
ta
by som
e
algorith
m
s. In
system
co
ntrol aspe
ct, ba
sed
on
re
cog
n
ition re
sult o
f
the image
d
a
ta, throu
gh t
h
e
external
co
nd
itioning
circuit
r
y , STM32
f
o
rm
s
st
and
ard
ind
u
stri
al si
gnal output. External
b
u
ttons
set the
corre
s
po
ndin
g
fun
c
tion
of the
corre
c
ti
ve p
o
s
ition
dete
c
tion
system
items. T
FT
screen
displ
a
y store
d
volume mat
e
rial info
rmati
on and ID of the co
rrective
positio
n dete
c
tor.
3. Acquisitio
n
of Corr
ec
tiv
e
Image
The co
rrectiv
e
image data
acq
u
isitio
n overall de
sig
n
d
i
agra
m
is sho
w
n in Figu
re
2.
Figure 2. Block
Diag
ram o
f
Image Data
Acqui
sition
Mast
e
r
c
ont
r
o
ller
colle
ct
s
colo
r ima
ge,
t
he c
o
lor i
m
ag
e in the
colo
r of each pixel
by the
R, G, and B
three
comp
o
nents d
e
ci
sio
n
, each
co
m
pone
nt (25
5
) the values
desirable,
su
ch a
pixel can have more than 1600 million (255 *
255 * 255) of the col
o
r
range. When master
controlle
r duri
ng pro
c
e
s
sin
g
, it costs too
much
co
mp
u
t
ation, thus we will co
nvert
color im
age
to
gray.
Colo
r i
m
age
co
nvert
s
to
gray ima
ge n
eed
ed
ca
lculate
the
effective lumi
na
nce
value
of t
he
image pixel, calcul
ating the
effective pixels with
gray value u
s
ed by
the formula (1) is a
s
follows:
Y
0
.3
RED
0.59G
REEN
0.11
B
L
U
E
(1)
4. Acquisitio
n
of Maximu
m Image Contras
t
Contrast i
s
a
measurement
whi
c
h
rep
r
e
s
ents
the
bri
g
htest white a
nd the d
a
rke
s
t bla
c
k
level of different brig
htne
ss bet
ween li
g
h
t and da
rk
area
s, the g
r
eater the
differen
c
e i
n
ra
n
g
e
rep
r
e
s
ent
s the contrast the
greate
r
, the smalle
r
the ra
nge of differe
nce
s
be
half the small
e
r [4]
.
The co
rrectiv
e
contrast of the image is
also
a refere
nce to the co
nce
p
t above. The co
re
function
of co
rre
ctive core functi
on
s
of the p
o
sition
d
e
tection
sy
stem is to id
ent
ify line and
e
dge
of co
rrective
image. If the
co
ntra
st b
e
twee
n lin
e of
corre
c
tive im
age
and
ba
ckgroun
d
colo
r i
s
small, or co
n
t
rast col
o
r of
both sid
e
s of
the
im
age
is
sm
all, thus it i
s
difficult to di
stinguish
them[5-6]. Be
cau
s
e
we will
use RGB three prim
ary
color LE
D to achieve the eff
e
ct of the lig
ht
mixing, we
re
sea
r
ch the
pl
acem
ent of
L
E
D. We u
s
e
d
pla
c
eme
n
t
manne
r
as Fi
gure
3
-
1
sh
o
w
s,
the RGB LE
D cro
ss
pla
c
ed in the fo
rmation of LE
D
dot mat
r
ix. It is feasible
according to
actual
light mixing effect. Chan
ge
the duty cycle
of thr
ee-way PWM wave
can formul
ate different light.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No. 4, April 2013 : 2124 – 2
130
2126
Figure 3. The
Placeme
n
t Way of Led Dot Matrix
We u
s
e the v
o
lume mate
ri
al pattern a
s
sho
w
n in Fi
g
u
re 4 a
s
exp
e
r
imental
subj
ects. Th
e
cha
r
a
c
teri
stic of the imag
e is a p
a
ttern of bl
a
ck lin
es o
n
a g
r
ay
backg
rou
nd.
Figure 5 is
the
volume mate
rial pattern
without light source
. The b
l
ack line on the pattern
cannot be fou
nd
promi
nent, a
nd
can
not b
e
identified.
T
h
rou
gh
di
st
ri
buting
differe
nt RGB
com
pone
nt color,
we
can
ge
nerate differe
nt li
ght. The fi
rst light pa
ttern a
s
we u
s
e duty
cycle
to qu
antify is
PWMRE
D
=1
0%, PWMGREEN=20%,
PWMBLUE=
50%. Th
e second
light pattern
is
PWMRE
D
=1
0%, PWMGREEN=2
0%, PWMBL
U
E=6
0
%
. U
nder the
first light pattern, co
rrecti
ve
strip
diag
ram
is
sho
w
n in
Figure 6. Un
der th
e seco
nd light p
a
ttern, corre
c
tive strip
diag
ra
m is
sho
w
n in Fi
g
u
re 7. Comp
are Fig
u
re 6 and Fig
u
re 7,
we ca
n say that different prop
ortio
n
s p
l
ay
importa
nt role
in black line i
dentificatio
n. We
can say that Figure 7 has a g
r
eate
r
contra
st.
Becau
s
e the
r
e are differe
nt targets an
d bac
kg
rou
n
d
s, different
scene
s nee
d
different
requi
rem
ents.
Target a
n
d
backg
rou
n
d
brightn
e
ss
contrast i
s
u
s
ually divide
d into appa
rent
contrast (so
m
e literature
called
the a
pparent co
ntrast), the inh
e
rent contra
st (also known as
zero co
ntra
st) and modul
ation co
ntra
st. Followin
g
we
will give their
definition
s
.
Modulatio
n
contra
st: the
ratio of diffe
re
nce
bet
wee
n
the target a
n
d
ba
ckgroun
d
to sum
of target an
d
brightn
e
ss o
f
the backg
ro
und b
r
ight
n
e
ss. Thi
s
d
e
finition is ofte
n use
d
when
the
black a
nd
whi
t
e grid te
st
CMOS sensor.
Definition
fo
rmula i
s
sho
w
n in Equ
a
tion
(2).
Wh
ere
M is
modulatio
n contra
st,
b
L
repre
s
ent
s the
lum
i
nan
ce
of the
target,
m
L
indicat
e
s th
e b
r
ightn
e
ss of th
e
backg
rou
nd.
mb
mb
L-
L
M
LL
(2)
Apparent con
t
rast: the ratio of the differenc
e bet
wee
n
target a
nd
backg
rou
nd b
r
ightne
ss
to backgroun
d brightn
e
ss, definiti
on formula sh
own in Equation (3
):
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Printing Dete
cting Algo
rith
m
basing on
Maxim
u
m
Degree of Reco
gnition (Hu Z
hang
)
2127
mb
b
L-
L
C
L
(3)
Inhere
n
t co
ntrast: In ap
parent cont
ra
st test,
whe
n
u
s
e the sky blu
e
as b
a
ckg
r
o
und, the
measured a
p
pare
n
t contra
st called i
nhe
rent
cont
ra
s
t. Definition formula is
s
hown in Equation
(4). W
h
e
r
e
0
C
is inhere
n
t cont
rast,
s
L
is the bri
ghtne
ss of th
e sky ba
ckground.
ms
0
s
L-
L
C
L
(4)
We a
c
qui
re
physi
cal ima
ge whi
c
h
co
ntains
the
b
ound
ary into
the microco
n
trolle
r,
throug
h digita
l image
pro
c
e
ssi
ng m
e
thod
s, dete
r
mine
colo
r
contrast
on both
si
de
s of the
bord
e
r.
We
al
so call both
sid
e
s of
the
border were colo
r
blo
c
ks,
we na
me
them col
o
r bl
ock 1 an
d col
o
r
block 2 re
sp
ectively. Whe
n
measure lo
cal co
ntra
st, we ca
n mea
s
ure gra
datio
n value of co
lor
block 1 an
d color blo
c
k 2 resp
ectively. Colo
r mod
u
la
tion contrast
of color bl
ock 1 and colo
r b
l
ock
2 are sho
w
n i
n
Equation (5
), also, the
av
erag
e co
ntra
st of corre
ctive
image:
Whe
r
e m, b are mea
s
ured gray scal
e va
lue of color blo
c
k 1 and colo
r block
2
r
e
spec
tively.
m-
b
M
mb
(5)
The contra
st
cal
c
ulatio
n method ab
ove also a
pplie
s to the cal
c
ula
t
ing the co
ntrast of
line
with the
backg
rou
nd
of co
rrective
i
m
age. T
h
e
brightne
ss of th
e light
source
is adju
s
ted
by
adju
s
ting th
e
PWM p
u
lse, the b
r
ightn
e
ss of the
LED a
d
juste
d
by
ch
angin
g
the
p
u
lse
duty. Th
e
three
com
p
o
nents
of R,
G, B determines the
di
fferent colors of visi
ble
light, and e
a
ch
comp
one
nt o
f
the 256 typ
e
s of valu
es.
So that
can
be a total range
of more than 1
6
mi
llion
colo
rs
(25
6
*2
56*25
6). Thi
s
not only increases
se
a
r
ch
time, and actually produ
ce su
ch a wi
d
e
variety of ligh
t, far beyond
the sp
eed
of 32-bit mi
cr
ocontrolle
r. The
r
efore,
we
sh
ould redu
ce t
h
e
variation
ra
n
ge of
color can b
e
d
r
a
w
n
to 2
50
kin
d
s of illumi
natio
n level.
We
need
find
out
an
optimum light
in the 250
ki
nds
of illumin
a
tion level,
which m
a
kes
maximum corrective
cont
ra
st of
the image. Search alg
o
rith
m is as follo
ws: [7-11]
1)
In the first ro
und
we
set
a
maximum lo
o
k
up
value
as
250, the
sm
al
lest valu
e a
s
0, step
ping
value is 26
=6
4, we can det
ermin
e
the scope of
the first rou
nd sam
p
les
R=[0, 63
, 126, 189,
252], G=[0,
63, 126, 189
, 252], B=[0, 63, 126, 18
9, 252]. There are 5 valu
es
for each
comp
one
nt, a total of 5*5*5 = 125 combi
nation
s
.
2)
Cal
c
ulate
co
ntrast of e
a
ch com
b
inatio
n,
there a
r
e
125 value
s
.
Comp
are the co
ntra
st
values,
cal
c
ulate the
ma
ximum contrast valu
e is
△
max1, co
rresp
ondi
ng to
the
sampl
e
values
are R=
R1, G=
G1, B=
B1.
3)
In the
se
co
nd
ro
und,
set
m
a
ximum
R i
s
R1
+64, mini
mum
lo
okup
value is R1
-6
4; maximum
G is G1
+6
4, minimum loo
k
up valu
e is
G1-6
4, maximum B is B1
+64, minim
u
m lookup val
ue
is B1-64, ste
pping value i
s
25
=32, sco
pe of
the sa
mples a
r
e R=[R1-6
4
, R1
-3
2, R1, R1+32
,
R1
+64], G
=
[
G
1-6
4
, G
1
-3
2, G1,
G1
+3
2, G1
+64
],
R=[B1
-
64, B
1
-32,
B1, B1
+32, B
1
+64],
determi
ne the
value of the element, discard the value
of which is le
ss tha
n
ze
ro;
4)
Cal
c
ulate co
ntrast of ea
ch combin
ation,
comp
are
the contra
st values, cal
c
ulate the
maximum contrast value i
s
△
max2, co
rresp
ondi
ng to the sample
values a
r
e R=R2, G=G
2
,
B=
B2
。
5)
In
the
thi
r
d ro
und, set
m
a
ximum R
i
s
R2
+32, minimu
m
loo
k
up
val
ue
i
s
R2-32; maximum G
is G2
+3
2, mi
nimum lo
oku
p
value i
s
G
2
-32, ma
ximum B is
B2+
32, mini
mum lo
oku
p
value is
B2-32,
step
pi
ng valu
e i
s
2
4
=1
6, sco
pe
of t
he sam
p
les are
R=[R2
-
32, R2
-16, R1, R2
+16,
R2+
32], G=
[G2-32, G2-16, G1, G2+
1
6, G2+
32], B=
[B2-32, B2-16, B1, B2+16, B2+
3
2],
determi
ne the
value of the element, discard the value
of which is le
ss tha
n
ze
ro;
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No. 4, April 2013 : 2124 – 2
130
2128
6)
Cal
c
ulate co
ntrast of ea
ch combin
ation,
comp
are
the contra
st values, cal
c
ulate the
maximum contrast value i
s
△
max3, correspon
ding to
the sample value
s
are R=R3, G=G3,
B=
B3
。
7)
Fourth
roun
d, fifth round, sixth round an
d
seventh ro
und u
s
e ste
p
p
ing value 2
3
=
8, 22=4,
21=2, 20=1 respe
c
tively, determi
ne a
sampl
e
value
for each ro
u
nd, com
pare
the cont
ra
st
unde
r ea
ch of
the set of sa
mple value
s
, thus
can get
maximum co
ntrast value o
f
R, G, B.
The ratio of RGB for sel
e
cted differe
nt light
is sho
w
n
in Table 1, percentag
e expre
ss a
s
the duty cycle
of the PWM wave which control
s
ea
ch
comp
one
nt.
Table 1. RGB Perce
n
tage
of 10 Kinds of
Light
Kind
Percen
1 2
3
4
5
6 7
8 9
10
R
10% 20%
30%
40%
50%
60% 70%
80% 90%
100%
G
10% 20%
30%
40%
50%
60% 70%
80% 90%
100%
B
10% 20%
30%
40%
50%
60% 70%
80% 90%
100%
Furthe
r, we
can get co
ntra
st value
of 10
flights as Ta
ble 2 sh
ows.
Table 2. The
Contrast Und
e
r Different Kinds Of Lig
h
t
Kind
Contras
1
2 3 4
5 6 7 8
9
10
△
0.9870
0.1879
0.2150
0.2990
0.3872
0.4990
0.5870
0.6550
0.7352
0.822
5. Recog
n
ition of Cor
r
ec
tiv
e
Image
Corre
c
tive p
o
sition
dete
c
tion system
is u
s
ed to
detect the
d
egre
e
of me
mbra
ne
deviation of the ho
rizontal
positio
n
in the cou
r
se of transmi
ssion. I
n
ord
e
r to det
ect the po
siti
on
of the offset, we sele
ct u
n
ique iconi
c image pa
tte
rn to track d
u
ring
co
rre
cti
v
e. Therefore,
recognitio
n
o
f
co
rre
ctive i
m
age
doe
s
not ne
ed to
recogni
ze
th
e entire im
a
ge, only vol
u
me
material p
a
ttern on the li
nes o
r
edg
es to be ident
ified and tra
cked. Re
solutio
n
of a corre
c
tive
image
coll
ect
ed by th
e m
a
ster controll
er is
320
*240,
the o
r
igin
al im
age i
s
565
RGB form
at, a
nd
then un
dergo
grad
ation p
r
oce
s
sing.
Corrective im
age
identificatio
n
is ba
se
d on t
he gray value
of
each id
entifie
d pixel. A
s
shown in
the
simulati
o
n
re
sult
colle
cted
co
rrective
i
m
age,
we
can
extract chara
c
teri
stics of co
rrective ima
ge [12]-[14].
Line featu
r
e
on corre
c
tive
image i
s
sho
w
n in
Fi
gu
re
8. Assuming
that the lo
cati
on of th
e
black lin
e is t
he 16
0th colu
mn and
the
width of
the bla
ck li
ne the
sa
me a
s
the
wi
dth of one
pixel
point. We ob
serve
pixel
s
i
n
one
row, from 1
s
t to 15
9th column,
p
i
xel gray valu
e is
not
cha
n
ge.
From 15
9 to 160 col
u
mn
and 160
-16
1
column the
gray value m
u
tate. From 161-320
colu
mn
,
gray valu
e
re
mains un
ch
a
nged.
The
fe
ature
of
co
rre
c
tive ima
ge
can b
e
rep
r
e
s
ented
as Fig
u
re
4-2, in the po
sition where the bl
a
ck lin
e unde
rgo
e
s p
o
s
itive mutation.
We
see
Figu
re 4
-
3, a
s
su
me si
de b
o
u
ndari
e
s
160t
h col
u
mn.
We ob
se
rve pi
xels in o
ne
row, fro
m
1st
to 159th col
u
mn, pixel gra
y
val
ue is not cha
nge. Fro
m
159 to 160
colum
n
and 1
6
0
-
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Printing Dete
cting Algo
rith
m
basing on
Maxim
u
m
Degree of Reco
gnition (Hu Z
hang
)
2129
161 colum
n
the gray value
mutate. Fro
m
161-
320
column, gray value remain
s
unchan
ged.
The
feature of corrective ima
g
e
can be
rep
r
ese
n
ted
a
s
F
i
gure 4
-
4, in
the positio
n whe
r
e the bl
ack
line und
erg
o
e
s
neg
ative mutation.
We can ide
n
tify and track a
c
co
rding
to th
is feature gray value mutation o
f
line o
r
edge.Assu
mi
ng ord
e
rs of
a corre
c
tive image graysc
ale a
s
Ta
ble 3 sh
ows.
Its resol
u
tio
n
is
320*2
40, we
use two-di
me
nsio
nal matri
x
to descript it as
320
2
4
0
320
240
[(
,
)
]
Ff
x
y
,
(,
)
f
xy
is the gray
value of (x,y), and
(
,
)
0
,
1
.
..,
255
fx
y
, 0-255 i
s
the total nu
mber of g
r
ay level of the image.
Table 3. Co
rrective Image
Gray Value A
rray
s
f(1,1)
f(1,2)
f(1,3)
f(1,4)
……
f(1,317)
f(1,318)
f(1,319)
f(1,320)
f(2,1)
f(2,2)
f(2,3)
f(2,4)
……
f(2,317)
f(2,318)
f(2,319)
f(2,320)
f(239,1)
f(239,2)
f(239,3)
f(239,4)
……
f(239,317
)
f(239,318
)
f(239,319
)
f(239,320
)
f(240,1)
f(240,2)
f(240,3)
f(240,4)
……
f(240,317
)
f(240,318
)
f(240,319
)
f(240,320
)
6. Results a
nd Conclu
sion
Figure 12
sh
ows the
corrective
pa
ge
of system, we ca
n se
e th
e co
rrective i
m
age a
nd
observe the a
c
tual po
sition
of the volume material line
or edg
e.
Figure 12. Co
rre
ctive P
age
Of This System
For
te
sting convenie
n
ce, assume
that the
edg
e is i
n
the cente
r
positio
n of t
he TFT
scree
n
wh
en the
volum
e
material
do
es
not deviati
on
, as sho
w
n i
n
the Fi
gu
re.
Obtain
point
s at
the edg
e a
s
marks, divid
e
the scre
en i
n
to 10 p
a
rts
i
n
the ho
rizon
t
al dire
ction.
Move the vol
u
me
material e
d
g
e
, make it coinci
dent wit
h
marks.
Wh
en the edg
e is coi
n
ci
dent
with the 1st
mark,
we me
asure
the output si
gnal voltage
VOUT of th
e
corre
c
tive p
o
sition d
e
tect
ion syste
m
b
y
a
multimeter. T
hen moving volume materi
al edge
ma
tch mark 2, mark 3 ......
mark 10 respectiv
e
ly,
measure
sign
al voltage V
O
UT at th
at tim
e
. Re
co
rd
th
e
ten g
r
ou
ps o
f
values, as shown in
Tabl
e
4 belo
w
.
Table 4. Bord
er Positio
n
of and Outp
ut Voltage
Boundar
y
position
0
1
2 3 4
5 6
7 8 9
10
VOUT
(V)
0.005
1.02
2.1
3.22 4.08
5.05 6.12
7.21 8.03 9.1
9.9
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No. 4, April 2013 : 2124 – 2
130
2130
Polyline grap
h is sh
own in Figure 13.
VOU
T(V)
0
2
4
6
8
10
12
01
2
3
4
567
89
1
0
边界位置
VO
U
T
(V
)
VOUT(V)
Figure 13. Polyline Grap
h Of Bounda
ry
Position And
The Outp
ut Voltage
From
ab
ove f
i
gure
s
we
ca
n say that
sy
stem
ca
n track the
e
dge
of
volume
mate
rial
and
has a go
od li
nearity. Setting the mark 5 is volume
mat
e
rial’
s
ce
nter
place, if VOUT(x)<VOUT(5
),
the volume
material offset to the left; if VOUT(x
)>VOUT(5),
the volume ma
t
e
rial offs
et to the
right. Acco
rdi
ng the val
ue
of VOUT
(x), t
he
cont
roller
can
control
the d
r
ive’s op
eration,
ch
an
ge
the lateral offset dista
n
ce of volume ma
terial, and thu
s
achieve correctio
n
effect.
Referen
ces
[1]
Pan Ru
oMi
ng.
T
he steel st
rip correctiv
e
cont
rol s
y
ste
m
. Harbin: m
a
ster'
s
degr
ee
thesis of
Northe
astern U
n
iversit
y
. 20
06;
1-10.
[2]
Niu Qin
g
Jun. S
t
rip correctio
n and co
ntrol s
y
s
t
em.
Non-ferro
us met
a
l proc
e
ssing.
20
03; 32
(3): 2-4.
[3]
Che
n
D
e
Ch
ua
n, T
ao Hon
gbi
n. A ne
w
t
y
pe
of co
rrective
c
ontrol s
y
stem
of
strip
w
i
nd
ing
process.
The
lear
ned j
our
nal
of Hang
z
h
o
u
Univers
i
ty of Electronic T
e
ch
n
o
lo
gy.
200
8; 28(2): 74-7
7
.
[4]
Yao Ju
nC
hoi,
Shi Ju
nsh
eng.
T
he explor
e
of color
ima
g
e
contrast d
e
fin
ed.
T
he
lear
ne
d jo
urna
l o
f
Yunn
an N
o
rma
l Univers
i
ty.
20
06; 26(4): 4
9
-5
1.
[5]
K W
enze, KL
a
dun
g AK S
a
mu
et. Measur
eme
n
t of Co
lor D
e
f
e
ctiv e
and
Nor
m
al C
o
lor
Visi
o
n Su
bjects
Color an
d
Lu
mina
nce
C
ontr
a
st
thresho
l
d F
unctions on CRT
.
Per iodica Lo ly t
e
chnica Ser Mec
h
Encino l.
2000
; 45(1): 103- 1
08.
[6]
KT
M ullen. T
he Contr ast S
ensitivit
y Of Hu
ma
n Col
o
r
Visio
n
T
o
Red Green And B
l
ue Ye
llo
w
Chromat icGrat
ing.
Journ
a
l o
f Physiolo
g
y.
1984; 35
9: 381-
382.
[7]
F
ang Ch
en
g. An improv
ed bi
n
a
r
y
se
arch a
l
g
o
rithm.
Moder
n
electron
ic tech
nol
ogy.
20
08; 5: 163-1
64.
[8]
Kunth DE. T
he Art of Computer Pr
ogram
ing,
3: Sorting an
d Search
ing.
Ad
diso
n W
e
sley.
197
3.
[9]
Santosa
PI. C
o
st an
d Be
nefi
t
of Informati
o
n
Se
arch
usin
g T
w
o
Differe
n
t
Strategies.
T
E
LKOMNIKA.
201
0; 8(3): 195
–20
6.
[10]
E Martiana, N
Ros
y
id, U Aguseti
a
. Docu
ment
Search
Engi
nee B
a
se
d on Automati
c Clusterin
g
.
TELKOMNIKA.
2010; 8(1): 4
1
–43.
[11]
H
y
afil L. Bou
n
d
s
for Selection.
SIAM Joural o
n
Co
mp
uterin
g
.
1976; 1: 109-
115.
[12]
Julez B. A Method of Co
din
g
T
V
Signals Ba
sed on Ed
ge D
e
tection.
Be
ll
tem
Te
ch
.C
omp
r
e
ssi
on
Vi
d
e
o
Television.
19
5
9
; 38(6R): 10
0
1
-10
20.
[13]
Larki
ao J, Mall
ykosk
i
P, N
y
l
a
nder
J. Predicti
on of Roll
in
g F
o
rce in Co
ldb
y
ph
ysic
al Mod
e
l
s
and Neur
al
computi
ng.
Jou
r
nals of Materi
als Process
i
ng
T
e
chno
logy.
1
996; 60: 3
81-3
85.
[14] Z
hou
Yi
Xi
ao.
Coating furnac
e strip
correction control system
.
T
a
iyuan:
Master'
s
degre
e
thesis o
f
T
a
iyua
n Univ
er
sit
y
. 20
08; 13-
18.
Evaluation Warning : The document was created with Spire.PDF for Python.