TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol. 14, No. 2, May 2015, pp. 318 ~ 32
2
DOI: 10.115
9
1
/telkomni
ka.
v
14i2.772
5
318
Re
cei
v
ed
Jan
uary 27, 201
5
;
Revi
sed Ma
rch 2
3
, 2015;
Acce
pted April 17, 2015
The Detection of Straight and Slant Wood Fiber through
Slop Angle Fiber Feature
Ratri D
w
i
Atmaja*, Er
w
i
n Susanto, Junartho
Halomoan, Gurni
t
a K I, Muhammad Ar
y
M
u
rti
Schoo
l of Elect
r
ical En
gin
eeri
ng, T
e
lkom Uni
v
ersit
y
Jala
n T
e
lekomunik
a
si no.
1,
T
e
rusa
n Bua
h
Batu, Band
ung
4
025
7, Indon
esi
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: ratrid
w
i
atm
a
j
a
@telk
o
mun
i
v
e
rsit
y
.
ac.i
d
A
b
st
r
a
ct
Quality co
ntrol
is one of i
m
p
o
r
t
ant process that can n
o
t be
avoi
ded i
n
in
du
stry. Image pro
c
essin
g
techni
qu
e is requir
ed to distin
guis
h
the qua
li
ty of
w
ood. If it can be do
ne
auto
m
atic
ally b
y
the comp
uter
, it
w
ill be very
he
lpful. This p
a
p
e
r discuss
es the d
e
tecti
on
of straight an
d s
l
ant w
ood fi
ber
to disting
u
is
h it
s
qua
lity. T
h
is paper pr
opos
es
an alg
o
rith
m
by usin
g onl
y
tw
o features i.e. me
an (aver
age va
lue
of slo
p
ang
le fiber) a
n
d
max
i
mu
ma
n
g
le (the
max
i
mum va
lu
e of sl
o
p
ang
le fiber).
T
hen the class
i
fication
meth
od
is
used by tresh
o
l
d
in
g. T
he resul
t
show
s the
performa
n
ce is ac
hiev
ed o
n
accu
racy 79.2%
Ke
y
w
ords
: det
ection, slo
p
an
gle fib
e
r featur
e, algor
ith
m
of feature extracti
on
Copy
right
©
2015 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
Image cl
assi
fication tech
nique
requi
res a fairly
long ste
p
. It starts from
image
segm
entation
,
object ident
ification,
feature extra
c
tio
n
, feature se
lection a
nd classificatio
n
[5].
Woo
d
image
real
-time se
g
m
entation alg
o
rithm ba
se
d
on video pro
c
e
ssi
ng ha
s
been p
r
op
osed
by Rat
r
i [1]
and
ha
s a
c
hi
eved 1
00%
accuracy. T
h
i
s
p
ape
r i
s
a
co
ntinuatio
n
of the
re
se
a
r
ch
pape
r [1] to distingui
sh the
straig
ht and
slant fiber
. The
sampl
e
s a
r
e
taken from previous
studie
s
on pap
er [1] and this p
ape
r only focu
s to the feature
extraction al
g
o
rithm.
2. Rese
arch
Metho
d
Figure 1 is th
e flowcha
r
t of algorithm p
r
opo
sed. Th
e sampl
e
s
are t
a
ke
n usi
ng webcame
and IP cam
e
ra. Bware
aop
en is u
s
e
d
to eliminate
the
noise (small
obje
c
ts) i
n
bi
nary imag
e. The
“bwl
abel
” i
s
u
s
ed
to find
co
nne
cted
obje
c
ts i
n
bi
nary i
m
age. Exam
ple of
bina
ry i
m
age ta
ke
n a
nd
output “b
wlab
el” re
sult
s is shown in Figu
re 2.
Figure 1. Flowchart of alg
o
rithm propo
sed
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
The Dete
ctio
n of Straight and Slant Wo
od Fibe
r Thro
ugh Slop Ang
l
e Fiber… (Ratri Dwi Atm
a
ja)
319
(a)
(b)
Figure 2.
(a)
Example of binary imag
e taken a
nd (b
)
O
u
tput bwla
bel
result
s from
Figure 2(a
)
Figure 3.
Flowchart to find
angle
an
d
da
ta
N from Figu
re 3 is the number of co
nn
ected obj
ect
s
in bwlabel re
sult. From Fi
gure 2
(
b
)
,
there a
r
e 3
conne
cted o
b
j
e
cts
so it get N=3.
Matrix
i
s
bina
ry imag
e of a con
n
e
c
ted obje
c
t. There
are 3
ma
tr
i
x
s
on Figu
re 2(b
)
.
Figure 4.
Third
ma
tr
i
x
from Figure 2(b
)
Area
i
s
the
a
m
ount of
1 va
lued pixel
in
ma
tr
i
x
. It is
obtained
area
=3 from Fi
gu
re 4. Th
en,
A
is 1-dim
e
n
s
ion
a
l matrix
which is
re
pre
s
ent
the
row po
sition
of 1 valued
pixel taken from
ma
tr
i
x
. while
B
is 1-dimen
s
ion
a
l matrix
whi
c
h i
s
re
prese
n
t the col
u
mn po
sition
of 1 valued pi
xel
taken from
ma
tr
i
x
. From Fi
gure
4, it is o
b
tained
A
=[2
1 2] then
B
=[
4 5 5]. It means that the
r
e
are
3 pixel
s
of 1
valued
pixel i
n
coordinate
(2,4),
(1,5
), a
nd (2,5).
Mi
n
is
the
lowest value
of
A
, while
ma
x
is the highe
st value o
f
A
. If
A
=[2 1
2] then obtain
ed
mi
n
=1 and
ma
x
=2.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 2, May 2015 : 318 – 322
320
(a)
(b)
Figure 5.
(a) Flowchart to find
leftrow
and
leftcolum
n
and (b
) Fl
owcha
r
t to find
right
row
a
nd
rightcolum
n
length
i
s
th
e
length
of
A
, If
A
=[2 1
2]
so it i
s
o
b
tain
ed
length
=3. Then,
ang
le
is
an
obje
c
t slop
e angle (wo
od fiber) and
can
be found
with
this equatio
n
:
tan
|
|
|
|
Data
is a 1-di
mensi
onal m
a
trix contain
s
the values o
f
angle
at a
woo
d
image. M
ean
is
the avera
ge
of
angle
value while
m
a
xi
m
u
m
angle
is the highe
st
angle
valu
e. These
m
ean
and
m
a
xim
u
m
angle
are u
s
ed
as feature
vector. Wh
erea
s,
the classificatio
n
use
s
tre
s
hol
ding
method. Tresholdin
g
is do
ne by the followin
g
rule
s:
a) If
ma
x
i
mu
ma
n
g
l
e
<
x
or
m
ean
<
y
, It is
deci
ded a
s
st
raight fibe
r
b) If
ma
x
i
mu
ma
n
g
l
e
>=
x
an
d
m
ean
>=
y
, It is deci
ded a
s
slant fiber
3. Results a
nd Analy
s
is
The sa
mple
s which are u
s
ed a
r
e take
n from wo
od
processin
g
indu
stry with the size
20cm x 8
c
m. Figure 6 is th
e sampl
e
:
(a)
(b)
Figure 6. (a)
Sample taken
through
web
c
ame a
nd (b) Sample take
n throug
h IP came
ra
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
The Dete
ctio
n of Straight and Slant Wo
od Fibe
r Thro
ugh Slop Ang
l
e Fiber… (Ratri Dwi Atm
a
ja)
321
To see the
perfo
rman
ce
of the propo
sed
algo
rith
m, it is teste
d
by u
s
ing t
he three
scena
rio
s
:
a)
The sa
mple
s
are taken by usin
g two types
of ca
mera
s to find out whi
c
h is the b
e
st.
b)
Do the optimi
z
ation of
x
to
find out the best value of
x
c)
Do the optimi
z
ation of
y
to
find out the best value of
y
Table 1. The
result of first scena
rio
No
T
y
pe of
Came
ra
Number of sam
p
les
T
o
tal Accur
a
cy
(
%
)
Straight fiber
Slant fiber
1
Webcam
592 w
oods
356 w
oods
948 w
oods
71.73
2
IP
Camera
150 w
oods
100 w
oods
250 w
oods
77.2
The expe
rim
ents in Tabl
e
1 is done wi
th a va
lue of
x = 4 and y
= 3.2 then, the results
sho
w
th
e a
c
curacy
usi
n
g
IP cam
e
ra
i
s
hi
ghe
r tha
n
web
c
ame.
These
re
sult
s a
r
e
used
a
s
a
referen
c
e for
usin
g the sa
mple of IP camera in the n
e
xt scen
a
rio.
Table 2. The
result of se
co
nd scen
ario
No Value
of
x
(degre
e
s) Accuracy
(%
)
1 2
77.2
2 3
77.2
3 4
77.2
4 10
77.6
5 15
74
6 20
66.8
The experi
m
ents
i
n
T
able
2
i
s
don
e wi
th
a
valu
e of y
= 3.2
the
n
, the
results sh
ow
th
at
optimal value
is on the value of x = 10.
In other
wo
rd
s, straig
ht wo
od fiber ha
s a value of slo
pe
angle fib
e
r b
e
low
10 d
egrees, a
nd for
slant fibe
r wo
od ha
s a val
ue of sl
op a
n
g
le fiber
beg
an
over 10 de
grees.
Table 3. The
result of third scena
rio
No Value
of
y
(degre
e
s) Accuracy
(%
)
1 2.2
68.4
2 3.2
77.6
3 3.7
78.4
4 4
79.2
5 4.2
78.4
6 4.7
74.8
7 5.2
72
The exp
e
rim
ents
of table
3 is do
ne
with a va
l
ue of
x = 1
0
the
n
,
the re
sult
s
show that
optimal value
is o
n
the val
ue of y = 4. In othe
r word
s, straight
wo
od fibe
r ha
s
an ave
r
age
valu
e
of slop
an
gle
fiber b
e
low 4
degree
s, an
d
for sl
ant
fibe
r woo
d
ha
s
an
avera
ge valu
e of sl
op a
ngl
e
fiber beg
an o
v
er 4 deg
ree
s
.
4. Conclusio
n
This
re
se
arch ha
s a
n
al
g
o
rithm
with 7
9
.2%
accu
ra
cy by u
s
ing
only two fe
ature
s
i.e.
m
ean
(avera
ge value
of sl
op an
gle fibe
r) and
m
a
xim
u
m
angle
(the
maximum val
ue of sl
op a
n
g
le
fiber). Thi
s
al
gorithm
can b
e
adapte
d
to other cases t
hat have a sa
me typical obj
ect.
Referen
ces
[1]
Ratri D
w
i At
maja. W
ood
i
m
age re
al-tim
e segm
entatio
n alg
o
rithm
b
a
sed
on vi
de
o proc
essin
g
.
Internatio
na
l Journ
a
l of Ima
g
i
ng & Rob
o
tics
. 201
4; 15(1).
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 14, No. 2, May 2015 : 318 – 322
322
[2]
Haral
i
ck, R
obe
rt M, Lin
d
a
G
Shap
iro. C
o
mp
uter a
n
d
Ro
bot
Visi
on
Vol
u
m
e
I. Ad
diso
n-W
e
sle
y
.
19
92
:
28-4
8
.
[3]
F
S
Najafa
ba
di,
H Pour
gh
ass
e
m
.
Surface a
nd C
o
rner
Def
e
ct Detectio
n
on T
ile Imag
es
Usin
g Gab
o
r
F
eatures, Lev
el Set Se
gme
n
tation
an
d D
o
t Product.
Int
e
rnati
ona
l Jo
u
r
nal
of Ima
g
in
g & Ro
botics
.
201
2; 8(2).
[4]
Ahmad
Nazr
i
Ali, Mo
hd Za
i
d
Ab
dul
la
h. One
Dim
e
n
sio
n
a
l W
i
th
D
y
n
a
m
ic F
eatur
es
Vector F
o
r Iris
Classific
a
tio
n
Using T
r
aditio
nal Su
pp
ort Vector Machi
nes.
Journ
a
l
of T
heoretic
al
and Ap
pli
e
d
Information T
e
chno
logy
. 2
014
; 70(1).
[5]
Nur Sh
az
w
a
ni
Kamaru
din, et
al. Comp
aris
on
Of Image Cla
ssificatio
n
T
e
chni
ques
Usin
g
Caltec
h 10
1
Dataset.
Journ
a
l of T
heoretic
al an
d App
lie
d Information T
e
chno
logy
. 2
015
; 71(1).
[6]
Nad
eem Mah
m
ood, et al. Image Se
gment
ation Meth
ods
a
nd Edg
e
Detec
t
ion: An Appl
ic
ation T
o
Knee
Joint Articu
lar
Cartila
ge E
d
g
e
Detectio
n.
Jo
urna
l of T
h
e
o
r
e
tical
an
d Ap
pl
ied Infor
m
atio
n
T
e
chno
lo
gy
.
201
5; 71(1).
[7]
Ning C
h
e
n
, Xi
ao-p
i
ng So
ng,
Yi Liu. Edge
Detection Ba
sed on Bi
omi
m
etic Pattern Reco
gniti
on
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2014; 1
2
(9).
[8]
Qu zho
ngsh
u
i.
An Al
gorithm
o
f
Image Qu
alit
y Assessme
n
t B
a
sed
on
Dat
a
F
i
tting of Ima
g
e
Histo
gram.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2014; 1
2
(1).
[9]
Hon
g
-an L
i
, Jie Z
hang, Ba
o
s
hen
g Kan
g
. Image D
e
forma
tion Base
d on
W
a
velet F
ilter and Co
ntro
l
Curves.
T
E
LK
OMNIKA Indon
esia
n Journ
a
l o
f
Electrical Eng
i
ne
erin
g.
201
4; 12(5).
Evaluation Warning : The document was created with Spire.PDF for Python.