TELKOM
NIKA
, Vol.14, No
.1, March 2
0
1
6
, pp. 254~2
6
1
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2486
254
Re
cei
v
ed Au
gust 23, 20
15
; Revi
sed
Jan
uary 13, 201
6
;
Accepte
d
Febru
a
ry 2, 20
16
Morphological Feature Extraction of Jabon’s Leaf
Seedling Pathogen using Microscopic Image
Melly
Br Ban
gun*
1
, Yeni Herdi
y
eni
2
,
Elis Nina Herli
y
ana
3
12
Department o
f
Computer Sci
ence Bo
gor Ag
ricu
ltura
l
Un
ive
r
sit
y
, Drama
ga
Bogor-In
d
o
nesi
a
3
Departme
n
t of Silvicult
u
re Bo
gor Agric
u
lt
ura
l
Universit
y
, Dr
amag
a Bog
o
r-Indo
nesi
a
*Corres
p
o
ndi
n
g
author, em
ail
:
mell
y_b
ang
un
@
y
ah
oo.co.id
1
,
y
e
ni.her
di
ye
ni
@ipb.
ac.id
2
,
elish
e
rl
ian
a
@
y
aho
o.com
3
A
b
st
r
a
ct
This research
aim
s
to analy
z
e morphologic
al techni
ques f
o
r feature extr
action
of Jabon’
s leaf
seed
lin
g p
a
tho
gen
usi
ng
di
gi
tal
microsc
opi
c i
m
ag
e.
T
h
e
kinds
of th
e
patho
ge
n w
e
r
e
Curv
ul
aria s
p
.,
Coll
etotrich
u
m
sp., and
F
u
sari
um sp.. Path
o
gens
or c
aus
e
s
of dis
eas
e w
e
re i
d
e
n
tified
ma
nu
ally
bas
e
d
o
n
macr
osco
pic a
nd microsc
opic
observati
on of
morp
ho
log
i
ca
l
characters. Morph
o
lo
gic
a
l ch
aracters descr
i
b
e
the ch
aracteris
t
ics of sh
ape, c
o
lor
an
d si
z
e
of a
patho
ge
n str
u
cture. W
e
foc
u
sed
on
sh
ape
feature
by
usi
n
g
the
mor
pho
lo
gi
cal tec
hni
qu
es
to featur
e extr
action. T
h
e
morph
o
lo
gy fe
atures
extraction
use
d
w
e
re
ar
ea,
peri
m
et
er, co
nvex
area,
c
onvex
p
e
ri
met
e
r, co
mp
actne
ss, soli
dity, c
onvex
ity, an
d
rou
ndn
ess.
T
he
meth
od
olo
g
i
e
s
w
e
re acq
u
isi
t
ion, pr
eproc
e
ssing, fe
ature
s
extractio
n
a
nd d
a
ta
ana
ly
sis for d
e
rivat
i
ve
features. W
i
th features extr
action, w
e
go
t the
pattern
that describ
ed
each p
a
tho
g
en for path
o
g
e
n
ide
n
tificatio
n
. F
r
om the
exp
e
ri
me
ntal
resu
lt s
how
ed
th
at co
mp
actness
an
d ro
und
ness
fe
ature w
e
re
a
b
l
e
t
o
differenti
a
te ea
ch patho
ge
n d
ue to
that the character
i
stics o
f
eac
h path
oge
n class w
e
re separ
ated.
Ke
y
w
ords
: fea
t
ure extraction,
Jabon,
micros
copic i
m
age, p
a
thog
en, morp
hol
ogic
a
l
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Jab
on (
Anth
oce
phal
us cadam
ba
(Ro
x
b.) Miq) is
a nativ
e Indone
sian fo
re
stry plant
widely
cultiva
t
ed be
cau
s
e
it has ma
ny advantag
es
over
othe
r plants.
It
is kno
w
n as a fast
gro
w
ing
sp
ecies [1-3], a cylindrical ro
d
with a go
od l
e
vel of align
m
ent rod
s
a
n
d
little bran
ch
ing
[2-4], able to gro
w
on vario
u
s types of soil, silvic
ultural treatment is relatively easy [1], and it is
one of the p
opula
r
altern
ative medici
n
a
l plant
s in
Indon
esi
a
re
cent years [3]. The seedli
n
g of
Jab
on h
a
s p
r
oblem
s with
dise
ases in t
he nu
rsery a
r
ea be
ca
use it
has th
e pote
n
tial to be a
host
for the patho
gen and
su
ccule
n
t seedli
ng con
d
ition
t
hat are still relatively vulnera
b
le [5].
The
dise
ase i
s
o
n
e
of th
e o
b
st
acle
s
be
cau
s
e it
can
redu
ce
qua
ntity a
nd q
uality. M
o
st di
se
ases
in
Jab
on
are
ca
use
d
by fun
g
i
[3]. One
sy
mptom o
r
di
sease
can
be
cau
s
e
d
by dif
f
erent p
a
thog
en
and the treat
ment is different either. Di
sea
s
e
s
and p
a
thoge
ns that
have been reporte
d to attack
Jab
on i
n
n
u
rseri
e
s in
clud
e leaf
spot di
sea
s
e
cau
s
e
d
by
Rhi
z
o
c
t
onia
s
p
.
[6] and
Colletot
r
ichum
sp. [7], leaf
blight
cau
s
ed
by
Fu
sa
rium
sp. [6]
and
dieba
ck di
se
ase
s
ca
used
by
Rhizocto
nia
sola
ni Kuhn.
[8] and
Botryodiplo
d
ia
sp. [5, 6]. The mos
t
s
i
gnific
ant
cha
r
a
c
t
e
ri
st
ics of
f
u
n
g
i t
o
be
identified a
r
e
spo
r
e
and
mycelium [9]
.
Spore i
s
o
ne imp
o
rtant
part in th
e
identificatio
n
of
morp
holo
g
ica
l
characte
risti
cs [10].
In this study path
ogen
s ca
me from
Deut
ero
m
ycete
s
c
l
ass
that are
impe
rfect fun
g
i b
e
c
au
se th
e o
n
l
y
kno
w
n
as a
namo
r
phi
c p
h
a
se
or a
s
exu
a
l pha
se,
so
to
identify it based on the
ch
ara
c
teri
stics of
asexual sp
ore
s
called
conidia [11]. T
he morphol
o
g
ical
feature of pat
hoge
n ca
n be
seen in Ta
bl
e 1 [5, 10, 12].
Re
sea
r
ch rel
a
ted to Jabo
n perfo
rme
d
by [13]
whi
c
h
identify the type of Ja
bon’
s fungi i
n
Sampali Me
d
an nu
rserie
s,
this re
se
arch
[5] was to
i
d
e
n
tify and test
the prim
ary causes
of fung
al
pathog
eni
city dieba
ck di
se
ase o
n
Ja
bon
seedli
ng.
Some research to
develo
p
pattern re
cog
n
ition a
n
d imag
e p
r
o
c
e
ssi
ng u
s
in
g digital
microsco
pic i
m
age
to d
o
[
14] an
auto
m
atic id
entificat
ion
a
nd cla
ssification of
Nosem
a
patho
gen
agent
s
with
segm
entation
tech
nique
s,
Scale Inva
ria
n
t Featu
r
e T
r
ansfo
rm
(SIFT) a
nd Su
pp
ort
Vector
Ma
chi
ne (SVM
). [15] to do mo
rp
hologi
cal
a
n
a
l
ysis o
n
the a
c
ute le
ukaem
ia identificatio
n
usin
g a microscopi
c imag
e of the blo
od syste
m
.
[16] to cond
u
c
t leukocyte
cla
ssifi
cation
for
leukaemia fo
r detectin
g
by
usin
g imag
e
pro
c
e
ssi
ng te
chni
que
s. [17
]
to do the ge
ometri
c featu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Morp
holo
g
ica
l
Feature E
x
traction of Jab
o
n
’
s Leaf
Seedling Pathog
e
n
usin
g… (M
elly Br Bang
u
n
)
255
extraction
of
Batik imag
e
usin
g Cardi
n
al Spli
ne Cu
rve
Rep
r
e
s
ent
ation.
[18] to do the featu
r
e
extraction a
n
d
cla
ssifi
catio
n
for multiple
spe
c
ie
s of Gyroda
ctylus E
c
topara
s
ite.
Re
sea
r
ch rel
a
ted patho
ge
n identificatio
n in Ja
b
on’
s leaf see
d
ling
is still a little to do
esp
e
ci
ally in comp
uter
sci
ence, this is
an opp
ortunit
y
to develop pattern recog
n
ition and im
age
pro
c
e
ssi
ng
o
n
it. Focus
of this
re
sea
r
ch
will
extra
c
t mo
rphol
og
ical featu
r
e
s
of Ja
bon’
s l
eaf
see
d
ling pat
h
ogen u
s
in
g di
gital microsco
pic imag
e.
Table 1. Morpholo
g
ical fe
ature of path
ogen
Pathogen
Morphological
features
Curvularia
sp.
Conidia Shape
Inequilateral
(h
a
v
ing unequal sides, i.e., a spor
e
that is flattened
on
one side),
ellipsoidal
(
e
llipti
cal in o
p
tical sect
ion)
mainly
4-
to 5-
cell w
i
th a septum c
entr
a
ll
y
located
C
o
lo
r
B
r
ow
n
Size
(17.5-
)18.7–2
3(-
30) x
8.7-12
.5(-
1
4
) µm
Colletotr
ichum
sp.
Conidia Shape
Oblong
(sho
rt-c
ylindrical, w
i
th
tru
n
cate ends of spores)
, cylindrical
(having parallel sides, like a
cy
linder
),
ovoid
(egg-sha
ped, i
.
e.,
narro
west at the
top of a solid)
1-
cell
Color
hyaline
(lacking color/transpar
ant)
Size
13-17.5 × 4.7
-
5.3
µm
Fus
a
ri
u
m
sp.
Macro
conidia
Shape
boat-shape/cano
e-shape
(having
shape like bo
at/ canoe)
, luna
te
(shaped like the ne
w
moon, cre
scentic)
, fusifor
m
(spindle-shap
ed
,
i.e., w
i
dest in the middle and narro
w
i
ng to
w
a
rd t
he ends),
allantoid
(slightly
curved,
w
i
th p
a
rallel walls and rounde
d ends, sausage-
shaped),
cylindrical
(having parallel sides, like a cylinder)
septate, mostl
y
4
-
celled
Color
hyaline
(lacking color/ transpa
rant)
Size
(17.5-
)
29.1
-
45 x 2.9-4.7
µm
Micro
conidia
Shape
Ellips
i
odal
(
e
llipt
i
cal in optica
l
sec
t
ion)
,
oblong
(s
h
o
rt-c
y
l
i
ndri
c
al
, w
i
th
truncate ends
of
spores), o
void
(
egg-shaped,
i.e., narr
o
w
e
st at t
h
e
top of a solid)
, cylindrical
(having
parallel sides, like a cy
linde
r)
, o
v
al
(broadl
y
elliptical)
, Fusiform
(sp
i
ndle-shaped, i.e., w
i
dest in the
middle and narr
o
w
i
ng to
ward th
e ends)
, reniform
(Kidne
y-shap
ed)
,
obovoid
(inversely ovoid, i.e.,attached at the smaller end)
Color
no septate, 1-celled
hyaline
(lacking color/ transpa
rant)
Size
6-15.8
x 1.9
-
3.7(
-
5
) µm
2. Rese
arch
Metho
d
Re
sea
r
ch m
e
thod comp
o
s
e
s
four
sta
ge.
The
r
e a
r
e data a
c
qui
sition, prepro
c
e
ssi
ng,
feature extra
c
tion and data
analysi
s
.
2.1. Data
Ac
quisition
Data a
c
q
u
isi
t
ion wa
s
co
ndu
cted in t
he La
bo
ratory of Entomo
logy De
part
m
ent of
Silviculture
Faculty of Fore
stry at Bogor
Agri
cultural Univ
ersity in Ja
nuary
-
Mei 2
015.
Microsco
pic i
m
age d
a
ta
wa
s taken u
s
ing
a mi
cro
s
cope
optila
b ca
mera an
d sto
r
ed i
n
JPG
format. Th
e
magnification
of the
mi
cro
s
cope
u
s
ed
t
he
sam
e
val
ue fo
r a
nalyzing
whi
c
h
ea
ch
image a
c
qui
si
tion pro
c
e
ss i
s
abo
ut 400x. The data
a
c
q
u
isition
stage
is sh
own in Figure 1.
Figure 1. Dat
a
acq
u
isitio
n stage (a) i
s
ol
ating of the pathoge
ns (b)
taking the p
a
thoge
nic tissu
e
(c) laying isol
ate in prep
aration (d
) acqu
is
ition imag
e usin
g optilab
digital micro
s
cop
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 254 – 2
6
1
256
The exampl
e of a micro
s
co
pic imag
e is s
hown in Figu
re 2. Before d
o
ing prep
ro
ce
ss, the
microsco
pic i
m
age is
crop
ping man
uall
y
and then we get sub
-
ima
ge.
Figure 2. manual crop
ping
(a) mi
cro
s
co
pic imag
e (b
) sub
-
ima
g
e
2.2. Preproc
essing
We
co
ndu
cte
d
a
se
rie
s
of
prep
ro
ce
ssin
g imag
e to
ge
t the be
st
se
g
m
entation
im
age th
at
will be extracted. The preprocessin
g stage can be seen in Figure
3. In this research, we used
150
sub
-
ima
ges f
r
om th
re
e patho
gen
s.
Sub-im
age
should
be
co
n
v
erted to
a g
r
ayscale
ima
ge,
then
we u
s
e
d
media
n
smo
o
thing. Th
e
smoothing
filter is u
s
ed fo
r
blurring
and
redu
cing
noi
se.
The me
dian fi
lter is
a
com
m
only u
s
ed
n
online
a
r o
p
e
r
ator that
repl
ace i
s
th
e o
r
iginal g
r
ay lev
e
l of
a pixel by the media
n
of the gray level
s
in t
he
pixels of spe
c
ified
neigh
borhoo
d [19, 20]. Thi
s
filter is often useful b
e
cau
s
e it can redu
ce noi
se
with
out blurring e
dge
s in the image. Moreo
v
er,
then
we
use
d
Otsu th
re
sholdin
g
, Otsu
method
is a
i
med at fin
d
i
ng the
optim
al value
for t
h
e
global th
re
sh
old [21]. It is based on th
e
intercl
a
ss va
rian
ce maxim
i
zation [22, 2
3
]. We appli
ed
region filling i
f
the image has
a hole
so that
it
can be solved. We
used
medi
an
smoothing again
as re
moval of
small detail
s
from an imag
e prio
r to (large) obj
ect extractio
n
, and b
r
idgin
g
of sm
all
gap
s in line
s
or cu
rve
s
[19, 21] and finall
y
we use
d
dil
a
te operation.
Figure 3. Pre
p
ro
ce
ssi
ng (a
) sub
-
ima
ge (b) grayscale (c) me
dian
sm
oothing (d) Ot
su
thresholdi
ng (e) fill hole (f)
medi
an
smo
o
t
hing (g
) dilat
e
operation
2.3. Featur
e Extrac
tion
Features of a
n
obj
ect
are
usu
a
lly u
s
ed
to
cla
s
sify the obj
ect. T
h
e
goal
is to transfo
rm
the imag
es i
n
to data
an
d
then to
extract info
rmati
on
refle
c
ting
the visual
p
a
ttern [1
6]. T
h
e
morp
holo
g
ica
l
feature
s
consi
s
t of ba
sic fe
atur
e
s
(ar
ea, pe
ri
meter,
conv
ex area,
co
nvex
perim
eter) and derivative
featur
es (compa
ctne
ss,
solidity,
convexity and
roun
dne
ss). Th
e
explanation
o
f
basic a
nd d
e
rivative feature
s
are as fo
llows:
The a
r
e
a
i
s
repre
s
e
n
ted
b
y
the total n
u
m
ber
of n
on-zero pixel
s
wi
thin the
boun
dary [24].
Area of a bin
a
ry regio
n
R
can be fou
n
d
by simply counting the im
age pixels th
at make up t
h
e
regio
n
[22].
Perimete
r (o
r circumfe
ren
c
e) of a
regi
on
R
is define
d
as
the
l
engt
h
of its oute
r
contour,
whe
r
e
R
mu
st be conne
cted [22]. The perimete
r
is
calculated b
y
measu
r
ing
the sum of t
h
e
distan
ce
s b
e
twee
n succe
s
sive bo
und
ary pixels [24].
The
simple
st
mea
s
ure of t
he pe
rimete
r
is
obtaine
d by counting the n
u
mbe
r
of bou
ndary
pixel
s
that belon
g to an obje
c
t [20].
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Morp
holo
g
ica
l
Feature E
x
traction of Jab
o
n
’
s Leaf
Seedling Pathog
e
n
usin
g… (M
elly Br Bang
u
n
)
257
The co
nvex area
is cal
c
ul
ating
the
con
v
ex
hull a
r
ea
in
whi
c
h th
e
empty a
r
ea
betwe
en
the co
nvex h
u
ll bou
nda
ry
and the
bou
ndary
obje
c
t,
loade
d obj
e
c
t and
the pi
xel values
that
inclu
ded in th
e obje
c
t area.
The co
nvex hull is the sm
allest polygo
n
conv
ex that contain
s
all points of the regio
n
R
[22]. The
convex perim
eter is t
he
circum
ference or li
mits on th
e convex hull.
T
he illustration of
basi
c
feature
s
is sho
w
n in
Figure 4.
A
re
a
C
on
v
e
x
h
u
ll
C
on
v
e
x
A
re
a
C
on
v
e
x
per
i
me
t
er
P
er
i
me
t
er
1
1
1
1
1
2
2
2
2
2
3
3
3
3
3
4
4
4
4
4
5
5
5
5
5
6
6
6
6
6
7
7
7
7
7
8
8
8
8
8
a
b
c
d
e
Figure 4. Basic feature
s
(a) area, (b
) pe
ri
meter,
(c) co
nvex hull, (d)
c
onvex area, (e)
convex
perim
eter
Comp
actn
ess
is
the relatio
n
between a regio
n
’s area
and
its
pe
ri
meter
[22].
A
c
cordi
ng
to [16], comp
actne
s
s is d
e
f
ined a
s
the
ratio bet
wee
n
the are
a
of
an obje
c
t an
d the area of
a
circle
with th
e sa
me p
e
ri
meter. Th
e
maximum va
lue of 1 to
form a
circle
. Comp
actne
s
s
cal
c
ulatio
n is
defined in Eq
uation (1
) a
s
belo
w
.
(
1
)
Rou
ndn
ess i
s
the
ratio
of
the a
r
ea
of
an obj
ect to
the a
r
ea
of a
circle
with t
he
same
perim
eter of the co
nvex hu
ll object [16]. Rou
ndn
ess calcul
ation is d
e
fined in Equ
a
tion (2
).
_
(
2
)
Solidity is the ratio of the
area
of an o
b
ject
to the a
r
ea of a
conv
ex hull of the object.
Solidity measure
s
the den
sity of an object [16]. So
lidity calcul
ation i
s
define
d
in Equation (3).
_
(
3
)
Convexity is the relative a
m
ount that a
n
obje
c
t differs fro
m
a co
nvex object, and thi
s
value re
presents the
ratio
of the
peri
m
eter of an
ob
ject’s
co
nvex hull to the p
e
rimete
r of the
obje
c
t itself [16]. Accordi
n
g to [25] the
convexit
y is defined
as t
he ratio of p
e
rimete
rs of
the
convex hull
with original
co
ntour. Conve
x
ity ca
lculatio
n is define
d
in Equation (4
) as bel
ow:
_
(
4
)
The illust
ration of derivative feat
ures is
shown in Figure 5.
Figure 5. Deri
vative features (a
) co
mpa
c
t
nes
s (b
) ro
un
dne
ss (
c
) soli
dity
(d) co
nv
e
x
ity
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 254 – 2
6
1
258
2.4. Data
An
aly
s
is
At this stag
e, we fo
cu
sed
on an
alyzin
g
der
ivative fe
ature
s
that
can differentia
te each
pathog
en. Th
e derivative feature
s
a
r
e compa
c
tne
ss,
solidity, conv
exity, and roundne
ss.
3. Results a
nd Analy
s
is
From all re
search metho
d
s, we analy
z
e erro
r of e
a
ch
sta
ge. In
data
acquisi
tion, we
have an erro
r beca
u
se we
did not focu
s
on acqui
siti
o
n
data stag
e, it can effect to identify each
pathog
en
correctly
so in t
he prep
ro
ce
ssing th
e seg
m
ented im
ag
e ha
s not p
r
eci
s
e
sha
pe
with
sub
-
ima
ge. T
he data ha
s erro
rs in th
e acq
u
isitio
n
and prepro
c
e
ssi
ng which are
sho
w
n
in
Figure 6.
Figure 6. Erro
r in data acqu
is
ition an
d prepro
c
e
s
sing
(a) su
b-im
age
and prep
ro
ce
ssi
ng re
sult o
f
data-7
(b)
su
b-ima
ge an
d prep
ro
ce
ssin
g result
of data-8 (c)
sub
-
i
m
age an
d prepro
c
e
s
sing
result
of data-14.
The analy
s
is
result of each
derivative feat
ure
s
patho
g
en are
sho
w
n
in Figure 7
-
1
1
.
a
b
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
02468
1
0
1
2
1
4
1
6
1
8
2
0
2
2
2
4
2
6
2
8
3
0
3
2
3
4
3
6
3
8
4
0
4
2
4
4
4
6
4
8
5
0
co
m
p
act
n
ess
da
t
a
co
lleto
tr
ichu
m
cu
r
v
ular
ia
fusa
r
i
um
Figure 7. Co
mpactn
ess fe
ature an
alysi
s
of eac
h pat
hoge
n (a
) dist
ribution d
a
ta of compa
c
tne
ss
feature (b)
co
mpactn
ess b
o
xplot
Comp
actn
ess is define
d
a
s
the ratio be
tween the
area of an obj
e
c
t and the a
r
ea of a
circle
with th
e same
peri
m
eter [1
6]. From Fig
u
re
7(a), we
can
se
e that
Curvul
aria
sp.
ha
s mor
e
circle
sha
pe
than othe
rs.
Almost all d
a
t
a of
each p
a
thoge
n can
discrimi
nate
and they h
a
v
e
simila
r
circle
sha
pe, b
u
t th
ere
are
som
e
patho
gen
s t
hat bel
ong
to
Coll
etotri
chu
m
sp. (data_
8,
data_2
7 and
data_4
0) g
o
t observed i
n
to
Fusa
rium
sp.
grou
p be
cau
s
e the
r
e a
r
e
a simila
r sha
pe
with
Fu
sa
r
i
um
s
p
. th
e sh
ap
e
s
ar
e
oblo
n
g
an
d
cylindri
c
al
and
data
distrib
u
tion
of
data_
8, d
a
ta
_27
and data_
40 are
i
n
Fu
sa
r
i
u
m
sp..
From
Figu
re 7(b
)
, we ca
n see
that
Colletotrichum
sp.
(dat
a_8,
data_2
7 a
n
d
data_
40) i
n
clu
de
as a
n
extrea
m
data. Based
on it
s
sm
all varia
n
ce, the
comp
actn
ess feature of
Fu
s
a
r
i
u
m
sp
. is
un
ifo
r
m. Be
s
i
de
s
,
Coll
e
t
otrichum
sp.
doe
s not
ha
ve
uniform d
a
ta
becau
se the
varian
ce of v
a
lue is
high
er than
Cu
r
v
ula
r
ia
sp. and
Fu
s
a
r
i
u
m
sp.
T
h
is
feature
can b
e
able to differentiate ea
ch
pathog
en.
Rou
ndn
ess i
s
the
ratio
of
the a
r
ea
of
an obj
ect to
the a
r
ea
of a
circle
with t
he
same
perim
eter of the convex hu
ll of the object
[16]. From F
i
gure 8
(
a
)
, we can see tha
t
Curvulari
a
sp.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Morp
holo
g
ica
l
Feature E
x
traction of Jab
o
n
’
s Leaf
Seedling Pathog
e
n
usin
g… (M
elly Br Bang
u
n
)
259
and
Coll
etotri
chum
sp. ha
ve more ci
rcular o
r
rou
n
d
shap
e than
Fusa
rium
sp.
,
almost all the
pathog
en dat
a are
rep
r
e
s
e
n
ted to ea
ch
pathog
en,
bu
t there are so
me patho
gen
s that belo
ng
to
Colletotric
hum
sp. (d
ata_
8, data_
27
a
nd d
a
ta_4
0)
got ob
se
rved
into
Fu
s
a
r
i
um
sp.
be
cau
s
e
there a
r
e
sim
ilar shap
e wit
h
Fusari
um
sp. the sha
p
e
s
are
ob
lo
ng
a
nd
cylindri
c
al
.
in Figure 8(b
)
we
can
see t
hat
Coll
etotri
chum
sp. (d
a
t
a_8, data_
2
7
and
data
_
4
0
) in
clu
d
e
s
a
n
extrea
m da
ta.
Based o
n
ro
undn
ess feature,
Coll
etotri
chum
sp. ha
s not uniform data beca
u
se the varia
n
ce
value is hig
h
e
r
than
Cu
rvul
aria
sp. and
Fu
s
a
r
i
u
m
s
p
..
a
b
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0
2
4
6
8
10
1
2
14
16
18
2
0
22
24
26
2
8
30
32
34
36
38
40
4
2
44
46
48
5
0
r
o
u
n
dn
ess
da
t
a
c
o
lle
to
t
r
ic
h
u
m
cu
r
v
u
l
a
r
ia
fu
sa
r
i
u
m
Figure 8. Rou
ndne
ss feature analysi
s
of eac
h patho
ge
n (a) di
strib
u
tion data of ro
undn
ess
feature (b) ro
undn
ess box
plot
a
b
0.
6
0.
6
5
0.
7
0.
7
5
0.
8
0.
8
5
0.
9
0.
9
5
1
0246
8
1
0
1
2
1
4
1
6
1
8
2
0
2
2
2
4
2
6
2
8
3
0
3
2
3
4
3
6
3
8
4
0
4
2
4
4
4
6
4
8
5
0
so
lid
it
y
da
ta
col
l
e
to
tr
ic
hum
cu
r
v
u
l
ar
ia
Figure 9. Solidity feature analysi
s
of each pathog
en (a) dist
ributio
n
data of solidi
t
y feature (b)
solidity boxpl
ot
Solidity is the
ratio
of the
area
of
an
o
b
ject
to
the
a
r
ea
of a
conv
ex hull
of the
obje
c
t.
Solidity measure
s
the
den
sity of an
obj
ect [16].
In t
h
is feat
ure,
a
r
ea
an
obje
c
t and it
s
conv
ex
area h
a
ve sig
n
ificant influe
nce if the values a
r
e
same
. From Figu
re
9(a), we ca
n see that almo
st
all data are a solid o
b
je
ct becau
se all
data seg
m
e
n
ted witho
u
t a hole in the
prep
ro
ce
ssi
ng
stage. A solidity value that gets ne
arer to 1
which mean
s a
solid o
b
je
ct without a h
o
l
e.
Colletotric
hum
sp. and
Cu
rvul
ari
a
sp. a
r
e so tightly that it gets di
fficult to distingui
sh bet
wee
n
the
two p
a
thog
en
s
it
ca
n b
e
se
en in
Fig
u
re
9(b
)
.
Fu
sa
riu
m
sp. do
es n
o
t have
unifo
rm data
du
e t
o
its
varian
ce val
u
e that is hig
h
e
r tha
n
the
ot
her
pat
ho
gen
.
The solidity feature may be
abl
e
to use
to
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 1, March 2
016 : 254 – 2
6
1
260
differentiate
pathog
en
Fu
sari
um
sp.
b
u
t it is ha
rd
to differentiat
e
path
oge
n
Colletotric
hum
sp.
and
Curv
ular
ia
sp., so solid
ity feature is not able to re
pre
s
ent the type of pathog
e
n
.
a
b
0.9
0.9
2
0.9
4
0.9
6
0.9
8
1
1.0
2
1.0
4
0
2
4
6
8
1
01
21
41
6
1
82
02
22
4
2
6
2
83
03
23
4
3
63
84
04
2
4
4
4
64
85
0
con
v
ex
i
t
y
dat
a
col
l
eto
t
r
i
ch
um
cu
rv
ular
i
a
fu
s
a
ri
u
m
Figure 10. Co
nvexity feature analys
i
s
of each patho
ge
n (a) di
strib
u
tion data of co
nvexity feature
(b) convexity
boxplot
Convexity is defined a
s
the ratio
of peri
m
eters of the convex hu
ll over that of the origi
n
a
l
conto
u
r [2
5]. In this featu
r
e
,
the peri
m
et
er of
a
n
obj
e
c
t and
its
con
v
ex perimete
r
have
significant
influen
ce if th
eir valu
e
are
equal.
The
convexity
feature
doe
s not
have a
unifo
rm data
du
e t
o
its
high va
rian
ce
.
Fusa
rium
sp
. doe
s not
ha
ve a unifo
rm
data be
ca
use
the vari
an
ce
value is great
er
than oth
e
rs
but ove
r
all from Fig
u
re
1
0
(b
)
we
c
a
n
se
e that
co
nvexity feature i
s
not a
b
l
e
to
differentiate e
a
ch
patho
gen
be
cau
s
e all pathog
en
s
ha
ve convex
sh
ape. Convexi
t
y is not a
b
le
to
rep
r
e
s
ent the
type of patho
gen, it
is cau
s
ed by thre
e types
of
pathog
ens
have spread alm
o
st th
e
same d
a
ta that would b
e
d
i
fficult to distingui
sh bet
we
en the three t
y
pes of patho
gen.
Comp
actn
ess and rou
ndn
ess feature
can b
e
use
d
to differentiate each pa
thogen
differentiate
becau
se thei
r varia
n
ce va
lue are
di
scri
minated. Th
e
varian
ce of
each feature
is
sho
w
n in Fig
u
re 11.
0
0.0
0
1
0.0
0
2
0.0
0
3
0.0
0
4
0.0
0
5
0.0
0
6
0.0
0
7
0.0
0
8
c
o
m
pac
tne
s
s
s
ol
i
d
i
t
y
c
on
vex
i
ty
r
o
und
n
e
ss
va
r
i
a
n
c
e
f
eatu
r
e
co
l
l
eto
t
ri
c
hum
cu
r
v
u
l
ari
a
fus
a
ri
um
Figure 11. Th
e varian
ce va
lue feature of
each patho
ge
n
4. Conclusio
n
This pa
per
pre
s
ente
d
to
analy
s
is m
o
rph
o
logi
cal
extraction
fe
ature
of
Jab
on’s leaf
see
d
ling
pat
hoge
n u
s
in
g
micro
s
copi
c image
s. T
h
e mo
rph
o
log
y
feature
co
nsi
s
ts
of ba
sic
feature
s
and
derivatives feature
s
. The
basi
c
features u
s
ed a
r
e
ar
ea, pe
rimet
e
r, convex area
and co
nvex perim
eter. Derivative
feat
ure
s
used
a
r
e
com
p
a
c
tn
ess, solidity, co
nvexity, and
roun
dne
ss. F
r
om th
e
re
sults, we
ca
n
co
ncl
ude
th
at de
rivative features (compa
ctne
ss
and
roun
dne
ss fe
ature
)
can be
cho
s
en to
differentiate
e
a
ch pathog
en.
Solidity feature is not a
b
le
to
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Morp
holo
g
ica
l
Feature E
x
traction of Jab
o
n
’
s Leaf
Seedling Pathog
e
n
usin
g… (M
elly Br Bang
u
n
)
261
rep
r
e
s
ent the
type of path
ogen, thi
s
i
s
cau
s
e
d
by th
e value
of pa
thogen
Coll
etotrich
um
s
p
.
a
nd
Curvul
ari
a
sp
. so tightly that it gets di
fficult to distingui
sh bet
wee
n
the two p
a
tho
gen
s, but ove
r
all
of data are solid witho
u
t a
hole. Co
nve
x
ity is not
able to rep
r
e
s
en
t the type of pathog
en, it is
cau
s
e
d
by three types of p
a
t
hoge
n have
spread al
mo
st the sam
e
data that wo
u
l
d be difficult
to
disting
u
ish betwee
n
the three
type
s of pathogen. O
t
her de
rivative feature ma
y be used to
get
more featu
r
e
repre
s
e
n
ted.
For the best
result, it
is nece
s
sary to add othe
r feature like texture
feature
or fu
sion of
several
feature. F
u
rt
her
st
udi
es
will be focused
on p
a
thoge
n
cla
ssifi
cation
or
identificatio
n
of Ja
bon’
s l
e
af se
edlin
g u
s
ing
mi
cro
scopic imag
es
without
cropp
ing
a
nd syste
m
s
can id
entify pathoge
ns of a
colony imag
e.
Referen
ces
[1
]
Kri
s
n
a
w
a
ti
H
,
Ka
l
l
i
o
M, Ka
nn
i
n
e
n
M. An
th
o
c
e
p
h
a
l
u
s
ca
da
mb
a
(Mi
q
.
): Eco
l
ogy
, Sil
v
i
c
u
l
tu
re
and
Productiv
i
t
y
. B
ogor: Ce
nter fo
r Internation
a
l
F
o
rest
y
Res
e
a
r
ch. 2011.
[2]
Mul
y
a
na D, Asmarahm
an C, F
ahmi I. Bertanam
ja
bon. Jak
a
rta: Agro Med
i
a Pustaka. 2
0
11.
[3]
W
a
risno, D
a
h
ana K. Pe
lu
an
g inv
e
stasi: ja
bon
ta
nam
an
ka
yu mas
a
de
pan. Jak
a
rta: Pener
bit PT
Gramedi
a Pustaka Utama. 20
11.
[4]
Dula
di. Pa
nd
u
an l
engk
ap
pe
mbibit
an j
a
b
o
n
(Anthoce
p
h
a
l
u
s cad
a
mba (
R
o
x
b.) Miq). B
ogor: IPB Pr.
201
3.
[5]
Aisah A
R
. Ide
n
tifikasi
dan
p
a
toge
nisit
a
s ce
nda
w
a
n
pe
n
y
e
bab
primer
pe
n
y
akit mati
pu
cuk pa
da
bib
i
t
jab
on (
Anth
oce
pha
lus ca
da
mb
a
(Ro
x
b.) Miq).
T
hesis. Bogor:
IPB; 2014.
[6]
Herli
y
a
n
a
EN,
Achmad, Ardi
ans
ya
h P. Pe
ngar
uh
p
u
p
u
k
orga
nik cair t
e
rha
dap
pertu
mbuh
an b
i
bit
jab
on (A
nthoc
eph
alus
ca
da
mba m
i
q.) d
a
n
keta
ha
na
nn
ya
terh
ad
ap
pen
ya
kit.
Jur
n
al S
ilvik
ultur
T
r
opika
. 20
12;
03(0
3
): 168-
17
3.
[7]
Anggr
ae
ni I. C
o
llet
o
trichum
s
p
. Pen
y
e
b
ab
Pen
y
ak
it Berc
ak D
aun
Pa
da
Beb
e
rap
a
B
i
b
i
t T
anama
n
Hutan D
i
Perse
m
aia
n
.
Mitra Hutan T
ana
man
.
2009; 4(1): 2
9
-
35.
[8]
Rahm
an MA,
Baksha MW
, Ahmed F
U
. Di
seases
and
pe
sts of tree spe
c
ies in f
o
rest n
u
rseries
an
d
pla
n
tatio
n
s in
Bang
lad
e
sh. D
haka: Ban
g
l
a
d
e
sh Agric
u
ltura
l
Rese
arch Co
uncil; 1
9
9
7
.
[9]
Agrios GN. Pla
n
t Patholo
g
y
. F
i
fth editio
n
. Ne
w
York: Els
e
vi
er Academ
ic Pr. 2005: 42
3.
[10]
W
a
tanab
e T
.
Pictorial
Atlas
of
Soil
a
nd S
eed
F
ung
i Mo
rpho
log
i
es
of
Cultur
ed F
u
ngi
an
d Ke
y t
o
Speci
e
s. Lon
d
on: CRC Pr. 19
94.
[11]
Herli
y
a
n
a
EN. Biod
iversitas d
an pote
n
si
ce
n
d
a
w
a
n
di In
don
esia. Bog
o
r: IPB Pr. 2014.
[12]
Barnett HL,
H
unter BB. Ilus
t
r
ated Gen
e
ra
of Imperfect F
ung
i.
Fourth e
d
itio
n
.
Minn
es
ota: Burg
ess
Publ
ishi
ng C
o
mpan
y. 19
98.
[13]
Purba MD. Id
entifikas
i
fung
i
pad
a pem
bib
i
tan ja
bo
n (
A
n
thoce
p
h
a
lus cada
mba
(Rox
b.) Miq) di
Sampa
li Med
a
n
. Scription. M
eda
n: Univ
ersitas Sumatera U
t
ara; 2013.
[14]
Ramos CMA, Nino E, Santo
s
M.
Automati
c classificati
on
of Nose
ma p
a
thog
en
ic age
nts throu
g
h
mac
h
i
ne visi
on
techniq
ues a
nd kern
el-b
ase
d
vector mac
h
i
nes
. Comp
utin
g Col
o
mbi
an
Confer
ence
.
201
3: 1-5.
[15] Scotti
F
.
Automatic
morp
hol
ogic
a
l a
n
a
l
ysis
for acute l
e
u
k
emia i
d
e
n
tific
a
tion
in
peri
p
hera
l
bl
oo
d
micr
oscop
e
i
m
ages
. Pr
oceedings
of IEEE Int
e
r
national Conf
er
enc
e on
Computational
Int
e
lligence f
o
r
Measur
ement
S
y
stems a
nd A
pplic
atio
ns. 20
05: 96-1
01.
[1
6
]
Pu
tzu
L
,
Ca
o
c
ci
G, D
i
Ru
b
e
rto
C
,
L
e
u
c
o
c
y
t
e
cl
a
ssi
fi
ca
tio
n
fo
r
l
e
u
k
a
e
mi
a
de
te
cti
o
n
u
s
i
n
g
i
m
age
process
i
ng tec
hni
ques.
Els
e
vier
. 2014.
[17]
F
anan
i A, Yu
ni
arti A, Suc
i
ati
N. Geometric
F
eature E
x
trac
tion
of
Batik I
m
age
Usi
ng
C
a
rdi
nal
Spl
i
n
e
Curve R
epres
e
n
tation.
TELK
OMNIKA.
2014
; 12(2): 397-4
0
4
.
[18]
Ali R, Hussa
in
A, Man M. F
eature Extracti
on
and C
l
assific
a
ti
on F
o
r Multi
p
le
Species
of G
y
rodact
y
l
u
s
Ectoparas
ite.
TELKOMNIKA.
201
5; 13(3): 50
3-51
1.
[19]
Gonzal
ez RC,
W
ood RE. Di
gi
ta
l imag
e pr
oc
essin
g
. T
h
ird editio
n
.
Ne
w
J
e
r
s
e
y
: Pe
arso
n
Pretice Ha
ll.
200
8.
[20]
W
u
Q, Merchant F
,
Castleman KR. Microsco
pe Im
age Proc
essin
g
. Ne
w
Y
o
rk: Elsevier A
c
adem
ic Pr.
200
8.
[21]
Herma
w
a
t
i
F
A
. Pengo
la
han ci
tra digita
l. Yog
y
ak
arta: Pen
e
rbit ANDI. 2013.
[22]
Burger W
,
Burge MJ. Princip
l
es of Digita
l
Image Proc
essin
g
-Core A
l
gor
ith
m
. Londo
n: Sp
ring
er. 200
9.
[23]
Otsu N. A
T
h
resho
l
d Se
lecti
on Me
tho
d
fro
m
Gra
y
-L
evel
Histogr
ams.
IEEE Trans. SMC
. 1979; 9(
1):
62–
66.
[24]
Saras
w
at M, A
r
y
a
KV, Autom
a
t
ed microsc
o
p
i
c imag
e an
al
ysis for leuk
oc
y
t
es ide
n
tificati
o
n
: A surve
y
.
Elsevier
. 2014:
20-33.
[25]
Yang M, Kpal
ma K, Ronsin J
.
A survey of shap
e feature e
x
traction techn
i
ques
. 20
08.
Evaluation Warning : The document was created with Spire.PDF for Python.