TELKOM
NIKA
, Vol.14, No
.2, June 20
16
, pp. 630~6
3
7
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v14i1.2675
630
Re
cei
v
ed Se
ptem
ber 11, 2015; Revi
se
d Jan
uary 20,
2016; Accept
ed Feb
r
ua
ry
8, 2016
Leaf Morphological Feature Extraction of Digital Image
Anthoc
ephalus Cadamba
Fuz
y
Yustika Manik*
1
, Yeni Herdi
y
eni
2
, Elis Nina Herli
y
ana
3
1,2
Department of Computer S
c
ienc
e,
Bogor Agricult
ural
U
n
i
v
ersit
y
,
Jl. Meranti, W
i
ng 20 L
e
ve
l 5, Darmag
a
, Bog
o
r 166
80
3
Departem
ent of Silviku
l
tur
,
Bogor Agr
i
cultur
al Univ
ersit
y
,
Jl. Meranti, W
i
ng 20 L
e
ve
l 5, Darmag
a
, Bog
o
r 166
80
*Corres
p
o
ndi
n
g
author, e- ma
il: fuz
y
.
y
ustik
a
@gmai
l
.com
1
, ye
ni.h
erdi
ye
n
i
@ipb.
ac.id
2
,
elish
e
rl
ian
a
@
y
aho
o.com
3
A
b
st
r
a
ct
T
h
is rese
arch i
m
p
l
e
m
e
n
ted
a
n
i
m
a
ge feat
ure ex
tractio
n
method usi
ng morph
o
lo
gic
a
l
te
chni
ques.
T
he g
oal
of th
is procc
e
ss
is
detec
ti
ng
ob
j
e
cts that ex
ist in t
he
i
m
ag
e.
T
he
imag
e is
conv
erted
int
o
a
graysca
le
i
m
a
ge for
m
at. Then, gr
ayscal
e
i
m
ag
e is
pr
o
c
essed
w
i
th tresho
ldi
ng
met
hod
to g
e
t i
n
i
t
ial
seg
m
e
n
tatio
n
. F
u
rthermore, i
m
a
ge fro
m
se
g
m
e
n
tatio
n
resu
lts are calc
ulat
ed usi
ng
morp
hol
ogic
a
l
meth
od
s
to find the ma
p
p
in
g of the orig
inal fe
atures in
to t
he new
features. T
h
is proc
ess is done to
get better clas
s
separ
ation. R
e
search co
nd
uc
ted on tw
o Antocep
hal
us
cad
a
mba (Ja
b
o
n
) l
eaf dise
ase
d
s
eed
lin
gs data
set
imag
e th
at co
n
t
aine
d l
eaf s
p
o
t
dise
ase
an
d
l
eaf b
lig
ht. T
h
e
resu
lts o
b
tain
ed
morph
o
l
ogi
cal fe
atures
su
ch
as rectan
gu
lar
i
ty, round
ness
,
compac
tnes
s, solid
ity, convex
ity, el
on
g
a
tion, a
nd
ec
centricity a
b
le
to
repres
ent the c
haracter
i
stic sh
ape
of
the symptoms
of the di
sease. All
pro
p
e
rties for
m
the
sympto
m
s c
a
n
be q
uantit
ativel
y expla
i
n
ed
by
the features f
o
rm.
So
it can
be us
ed to re
p
r
es
ent type
of sympto
m
s
of tw
o
dise
ases in A
n
tocep
hal
us cad
a
mba (Ja
b
o
n
).
Ke
y
w
ords
: ant
ocep
hal
us cad
a
mba, feat
ur
e extraction,
mor
pho
logy
Copy
right
©
2016 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Anthoce
phal
u
s
ca
damb
a
(Jab
on) i
s
a type of
comm
erci
al plantati
ons of fast
-g
rowi
ng
local
peo
ple
(fast g
r
o
w
ing
spe
c
ie
s) an
d ca
n g
r
o
w
well o
n
the
acreag
e u
s
e
d
for
cultivation
,
shrub
s
, and swamp forest
s are wide
sp
re
ad in forest a
r
ea
s in Indon
esia. Ja
bon
can be used for
reforestatio
n and affore
sta
t
ion in orde
r to in
crea
se
produ
ctivity
of the land, and sh
ould
be
develop
ed in
indu
strial fore
st plantation
s
, as deman
d for wo
od is in
crea
sing [1].
Plantation de
velopment th
at has i
m
pli
c
ations
fo
r the
kind
of tree
planting
(mo
n
o
cultu
r
e
)
on a la
rge
scale re
quires t
he availability
of high-
qualit
y seed
s in
su
fficient qua
ntities [2]. On th
e
other ha
nd this tren
d imp
a
ct on the e
m
erg
e
n
c
e
of
the disea
s
e
[3] that causes ha
rm, am
ong
others, redu
ces th
e qu
anti
t
y and qu
ality of the re
sults
a
s
well as
i
n
crea
se pro
d
u
ction co
st
s
[4].
Acco
rdi
ng to
Angg
rae
n
i a
nd
Wibo
wo
[5] the
su
cc
e
s
s of fo
re
st
plantation
de
velopment
st
art
s
from se
edlin
g
s
pro
d
u
c
ed from the nurse
ry.
Duri
ng thi
s
p
e
riod
leaf di
sea
s
e
s
re
cei
v
e less atten
t
ion be
cau
s
e
they do n
o
t ca
u
s
e
signifi
cant l
o
sse
s
, except o
n
the
se
edlin
gs i
n
th
e
nu
rsery.
Ja
bon
l
eaf di
sea
s
e
that attacks t
h
e
nurse
ry pha
se in Bogor
re
ported by
Herliyana
et al.
[6] and Aisa
h [7], namely dieba
ck, leaf
spot
dise
ase an
d l
eaf blight. T
h
ese
thre
e di
seases ar
e
ca
use
d
by fun
g
i
.
Fungi
ca
use
local sym
p
to
ms
or
systemi
c
symptoms i
n
it
s h
o
st.
Gen
e
rally, fungi
cau
s
e l
o
cal n
e
crosi
s
, tissue
n
e
crosi
s
com
m
on
or kill plants [8].
Symptoms a
nd sig
n
s of d
i
sea
s
e h
a
ve an impo
rt
ant role to diag
n
o
se the di
se
a
s
e. With
sy
mpt
o
m
s
d
a
n
si
gn
s,
we
can
al
so
det
ermin
e
m
o
rp
hology and
chara
c
te
rist
i
c
of
the ca
usative
pathog
en. Experts
can
ide
n
tify the type of di
sea
s
e ba
s
e
d
on
th
e
vis
i
b
l
e
s
y
mp
to
ms a
n
d
s
i
gn
s
.
With image p
r
ocessin
g
techniqu
es, the image of sy
m
p
toms an
d si
gns is p
r
o
c
e
s
sed to obtain
the
cha
r
a
c
teri
stics fo
r id
entification p
r
o
c
e
s
s. On
Anth
oce
phalu
s
Cada
mba
(Jabo
n) plant
se
edli
ngs,
the blot
che
s
on leave
s
(le
a
f sp
ot) a
r
e i
ndicated t
hat
leaf ha
s di
se
ase
sympto
m
s
. The
leaf
sp
ot i
s
the death
of
necroti
cs th
at have
sha
r
p margin
s a
nd
it is th
e
result of lo
cal infectio
n b
y
a
pathog
en. Th
e colo
r of spo
t
s is colo
red f
r
om ye
llow to
brownish. The spots
can e
i
ther be ro
un
d-
sha
ped, oval
-sha
ped
or
sh
apele
s
s. If the sh
ape i
s
ro
und, the di
se
ase i
s
called l
eaf sp
ot dise
ase.
If spotting or
death o
c
curre
d
rapidly in whole or
in p
a
rt
, the disea
s
e
is call
ed bligh
t
[9].
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Leaf Morphol
ogical Featu
r
e Extra
c
tion o
f
Digital Im
ag
e Anthoce
p
h
a
lus… (Fu
z
y Yustika M
ani
k)
631
Digital ima
g
e
pro
c
e
s
sing
techni
que
s today have
gro
w
n ve
ry rapidly
with
a fairly
extensive ap
plicatio
n in variou
s fields. In image p
r
o
c
essing, in ord
e
r to make the pro
c
e
ss of
with
dra
w
al of info
rmation o
r
de
scription of th
e objec
t o
r
object re
co
gnit
i
on that exists in the imag
e,
feature
extra
c
tion pro
c
e
ss
is re
quired. Feature
extra
c
tion i
s
th
e p
r
oce
s
s fo
r find
ing the
ma
ppi
ng
of origi
nal fe
ature
s
to
ne
w featu
r
e
s
in
whi
c
h
it is
e
x
pected
to re
sult in
better
cla
s
s sepa
rat
i
on
[10]. Feature
extraction
is an impo
rtan
t step in
the
cla
ssifi
cation
[11], beca
u
se well
-extra
ct
e
d
features will
be able to increase the lev
e
l of accu
racy, while features that
are not well-extracted
will tend to exace
r
bate the l
e
vel of accuracy [12].
T
o
ide
n
t
ify o
r
c
l
ass
i
fy o
b
j
ec
ts
in
th
e imag
e
,
firs
t we
mu
s
t
e
x
tra
c
t fea
t
u
r
es
fro
m
an
image
and then u
s
e this feature in
a pattern to obtain a fi
nal grad
e cla
s
sifier. Feature extraction is u
s
ed
to identify feature
s
that can ma
ke a
b
e
tter re
pre
s
e
n
tation of th
e obje
c
t. Not
only col
o
r a
n
d
texture, shap
e or morphol
ogy can al
so be use
d
to
extract features. Mathematical morph
o
log
y
is
a tool to
extract im
age
co
mpone
nts th
at are u
s
ef
ul
to re
pre
s
e
n
t
and
de
scribe
sh
ape
of
reg
i
on,
su
ch a
s
bou
ndari
e
s, skel
etons, and t
he convex h
u
ll. Morphol
o
g
ical is
relat
ed to certai
n
operation
s
th
at are
u
s
eful
for an
alyzin
g
the shap
e of
the imag
e
so
that shap
e o
f
the obj
ect
ca
n
b
e
r
e
c
o
gn
ized
.
Zinove
et
al.
[13] ha
s
pre
d
icted
the
ra
diologi
st
a
ssessment
of L
ung Ima
ge
Databa
se
Con
s
o
r
tium
(LIDC) n
odul
e
s
u
s
in
g 6
4
f
eature
ima
g
e
s
of th
e fou
r
cate
gori
e
s (forms, i
n
ten
s
i
t
y,
texture, and
size). P
u
tzu
et al.
[14] ha
s
b
een d
o
ing
re
search fo
r id
e
n
tifying and
classifying whi
t
e
blood
cells (l
eukocyte
s
)
b
a
se
d o
n
mo
rpholo
g
ic
al
fe
ature
s
,
colo
rs a
nd textu
r
e
s
. Ga
rtne
r
et
al.
,
[15] has used
morphol
ogi
cal feature
s
su
ch a
s
ro
u
ndn
ess and elo
n
gation to
cla
s
sify zircon grains
from s
e
diment.
The proble
m
in this re
search i
s
ho
w to
identify feature
s
that can ma
ke a goo
d
rep
r
e
s
entatio
n of the
obj
e
c
t ba
se
d o
n
i
t
s shap
e. So
,
it ca
n b
e
used to fin
d
sig
n
ificant fe
atures
area in the im
age.
2. Rese
arch
Metho
d
2.1. Data Se
t
The data
use
d
are th
e im
age of two ty
pes
of
leaves Jab
on affect
ed by leaf di
sea
s
e
s
.
The di
sea
s
e
s
that are focu
sed in thi
s
re
sea
r
ch ar
e le
af spot an
d leaf blight ± 4
months. Data in
the
form
of symptomati
c leaves Ja
bon
obtaine
d fro
m
the ob
se
rvation of the
symptom
s
a
nd
makin
g
Ja
bo
n example
s
that sho
w
sy
mptoms of le
af spot and l
eaf blight
of 2 nurse
ry locations
arou
nd the campu
s
of Dra
m
aga a
r
e on
e nurse
ry located in Situ Gede a
r
ea a
nd one nu
rse
r
y in
IPB show in
Figure 1. Th
e numb
e
r of
plant samp
le
s taken fro
m
the location
of the nurse
ry is
adju
s
ted to the nurse
ry con
d
ition.
Figure 1. Dat
a
colle
ction
si
tes
2.2. Methodo
log
y
Methodol
ogy in gene
ral ca
n be de
scribe
d in Figure 2.
Figure 2. Methodol
ogy
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 630 – 63
7
632
2.2.1. Image Acquisi
tion
The Sympto
matic leaf ph
otogra
ph was taken by u
s
i
ng a digital camera for ev
ery kin
d
of
dise
ase. Example of this p
hoto is sho
w
n
in Figure 3.
Leaf Sp
ot Di
sease: Sympto
ms
and
si
gn
s of l
eaf
spot d
i
sea
s
e
are
ge
nerally th
e
sa
me o
n
each affe
cte
d
pla
n
ts,
whi
c
h
are
sores or blemi
s
h
e
s
that
are lo
cal to
the
ho
st leaf
co
nsi
s
ts of
dead
cell
s (n
ecrosi
s) in th
e leaves [8].
The a
r
ea of
necro
sis
ran
g
i
ng from
sma
ll to large
with a
form of i
r
regu
lar u
n
til unifo
rm [16]. Symptoms of
l
eaf
spot di
sea
s
e
Jabon
seed
s
can b
e
seen
i
n
Figure 3(a
)
.
Leaf Blight
Disea
s
e: Symp
toms
and
si
g
n
s th
at ha
pp
en in
the
org
an le
aves,
branche
s,
twigs
and flo
w
ers tu
rn b
r
o
w
n very q
u
ickly and t
horou
ghly are th
e causes
of deat
h [8]. There a
r
e
spot
s on the leaves op
aqu
e, dark b
r
o
w
n surrou
nde
d
by a chloroti
c halo [9]. Symptoms of le
af
spot di
sea
s
e
Jab
on seed
s
can b
e
se
en i
n
Figure 3(b
)
.
(a)
(b)
Figure 3. The
image of the dise
ased leaf
(a) Le
af spot
disea
s
e (b) L
eaf blight disease
2.2.2. Prepro
cessing
At prep
ro
ce
ss
stage, the
r
e
are
two
step
s to p
r
o
c
e
s
s t
he
captu
r
ed
image. T
he p
r
oce
s
se
s
are
cutting
off (croppi
ng) a
nd segme
n
ta
tion the im
ag
e. Croppin
g
tech
niqu
e is
d
one to
cut an
d
take di
sea
s
e
s
pa
rt in the image
s, just
like sho
w
n b
e
low (Figu
r
e
4(a
)). Mea
n
while, the obje
c
t
image
se
gme
n
tation p
r
o
c
e
s
s is do
ne to
obtain
a bin
a
r
y imag
e of t
he o
b
je
ct ima
ge by
usi
ng t
he
con
c
e
p
t of m
o
rph
o
logy
co
nsi
s
ts
of thre
shol
di
ng,
ed
ge d
e
tectio
n, and
op
enin
g
. A process of
feature extra
c
tion is foll
owed after the
pro
c
e
ss
of image segm
ent
ation [17]. Segmentation i
s
a
very importa
n
t
step in obje
c
t re
cog
n
ition
.
Ther
e a
r
e v
a
riou
s
segm
e
n
tation metho
d
s that can b
e
us
ed [18].
(a)
(b)
Figure 4 Prep
rocessin
g (a
) Cro
ppin
g
(b
) Segmentatio
n
As sho
w
n in
Figure 4(b
)
, Thre sholdi
ng
pro
c
e
ss b
a
se
d on Otsu me
thod wa
s app
lied to
the original images
[12]. Process
is
conti
nued by objec
t edge
detec
tion us
ing canny edge
detection technique to
get
the e
dge lines
of the obj
ect that
w
ill
be used to
calcul
ate objects
perim
eter fea
t
ures [12]. 'Holes' th
at are
f
ound in th
e
binary im
ag
e obje
c
ts fro
m
applying t
he
segm
entation
pro
c
e
ss i
s
fill
ed with
applyi
ng the
mo
rp
h
o
logi
cal op
eni
ng process
so it become
s
a
fully binary image obj
ect area.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Leaf Morphol
ogical Featu
r
e Extra
c
tion o
f
Digital Im
ag
e Anthoce
p
h
a
lus… (Fu
z
y Yustika M
ani
k)
633
2.2.3. Morph
o
logical Feature Extra
c
tio
n
To re
co
gni
ze
an obj
ect in t
he imag
e, so
me f
eatures
must b
e
extra
c
ted first. Morpholo
g
y
of the di
gital i
m
age i
s
th
e f
a
ct that
digita
l image
conta
i
ning
se
rie
s
o
f
pixels th
at
make
colle
ction
of two
-
dime
n
s
ion
a
l d
a
ta.
Certai
n m
a
th
equ
ation
s
o
n
a
se
rie
s
of
pixels
ca
n b
e
used to
imp
r
ove
asp
e
ct
s of the form and
structu
r
e
so
it can be ea
sie
r
to recogni
ze.
There a
r
e
several fe
atures
of the
sh
ape th
at can
be
cal
c
ul
ate
d
, like
a
r
ea
whi
c
h i
s
cal
c
ulate
d
b
a
se
d on the
numbe
r of pixels t
hat o
c
cupie
s
the
obje
c
t image
, the perime
t
er
(bou
nda
ry ob
ject) i
s
calcul
ated ba
se
d o
n
the
num
be
r of pixels a
r
o
und the
obje
c
t image. Based
on a
r
ea
an
d
perim
eter fe
a
t
ures,
value
s
of other
mo
rp
hology fe
atures
ca
n al
so
b
e
calculated
as
well. The foll
owin
g eq
uations
are fo
rm
ulas th
at
ca
n
be u
s
ed to
extract mo
rp
hologi
cal feat
ure
s
[13-15].
Table 1. Fo
rmula Morphol
ogical Featu
r
es
Morphological
Features
Formul
a
Descr
iption
Rectangularit
y
ar
ea
major
axis
xminor
axis
Technique to illustrate similarity
of object shap
e
w
i
th
rectangular shap
e.
Elongation
1
minoraxis
major
axis
Measuring the le
ngth of the object
.
Solidity
ar
ea
con
v
e
x
_
ar
ea
Measuring the d
ensity
of t
he obj
ect, ratio of area
to the
full convex object.
Roundness
4x
π
x
ar
e
a
con
v
e
x
_
perimet
e
r
Technique to illu
strate the level o
f
determination object.
Value 1 for circular object
is greater than one for not
circular object.
Convexit
y
con
v
e
x
_
perimet
e
r
parimet
e
r
The r
e
lative amount that
the o
b
ject is different fro
m
convex hull. This value
is the p
e
rimeter ratio b
e
tween
convex full of obj
ect and the
obje
c
t itself. If the value is1,
the object is called convex hull,
w
h
en the value i
s
bigger
than 1 the object is not convex hull or object w
i
th
irregular bo
undar
ies.
Compactness
4x
π
x
ar
e
a
perimet
e
r
The ratio bet
w
e
e
n
the object
area w
i
th circle area uses
the same perime
t
er.
Eccentr
i
city
majo
r
axis
m
inor
axis
major
axis
The ratio
of distance bet
w
een t
h
e
ellipse focal and major
axis. The value is betw
e
en 0-
1.
This resea
r
ch
is based o
n
analysi
s
u
s
in
g symp
tomi
c
spot
s’ sh
ape
of leaf spot di
sea
s
e.
The extracte
d shap
e feat
ure i
s
a
featu
r
e that
ha
s n
u
meri
cal
data
as
sho
w
n in
Figure 5. A
r
e
a
i
s
the wide
of spots (Figu
r
e
5(b
)), pe
rimet
e
r is th
e
pe
ro
phery o
r
limit
of spots
(Fig
ure 5
(
c)), maj
o
r
axis i
s
the
le
ngth of
sp
ots mea
s
u
r
ed
from the
ba
se
of the le
af to
the tip, mea
n
-
whil
e mi
nor
axis
is the width o
f
spots mea
s
ured from the
widest l
eaf surface (Fig
ure 5d),
the co
nvex hull (Fig
ure
5(e
)), convex area
(Figu
r
e
5(f)), and
the
convex pe
rim
e
ter (Fi
gure 5
(
g)).
(a)
(b)
(c
)
(d)
(e)
(f)
(g)
Figure 5. (a)
The ori
g
inal i
m
age, (b
) Are
a
, (c
) Pe
rimet
e
r, (d
) Mayor
Axis and Min
o
r Axis,
(e) Convex Hull, (f) Conv
ex
Area, (g)
Convex Peri
meter
The feature is also use
d
to calculate
round
ne
ss, solidity,
elon
gat
ion, ecce
ntricity,
comp
actn
ess, convexity, a
nd
re
ct
angul
a
r
ity. Illustratio
n
for all fe
atu
r
es
can
be
seen i
n
Fi
gure
6
belo
w
:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 630 – 63
7
634
(a)
(b)
(c
)
(d)
(e)
(f)
(g)
Figure 6. (a)
Rou
ndn
ess, (b) Solidity, (c) E
ccent
ricity
,
(d) Conv
i
c
ity
,
(e) Comp
actn
es
s,
(f) Elongatio
n, (g) Recta
n
gularity
3. Results a
nd Analy
s
is
Table 2. Feat
ure
s
Extractio
n
Leaf
Diseases
Jabon
S
y
mptoms
image
Features E
x
tracti
on
Roundness
Rectangularit
y
Compactness
Convicity
Solidity
Elongation
Eccentr
i
city
Leaf Spot
Disease
0.506
0.738
0.733
1.203
0.918
0.157
0.537
0.456
0.781
0.747
1.280
0.951
0.142
0.514
0.523
0.744
0.734
1.185
0.933
0.010
0.142
0.516
0.765
0.824
1.263
0.952
0.016
0.180
0.539
0.748
0.787
1.208
0.952
0.058
0.337
0.570
0.786
0.829
1.206
0.964
0.099
0.434
0.475
0.793
0.751
1.258
0.952
0.008
0.122
0.431
0.790
0.762
1.329
0.961
0.189
0.586
0.439
0.805
0.766
1.321
0.932
0.032
0.251
Leaf Blight
Diseases
0.172
0.369
0.169
0.497
1.049
0.626
0.927
0.204
0.480
0.198
0.636
0.910
0.778
0.975
0.208
0.523
0.149
0.731
1.069
0.688
0.950
0.197
0.348
0.188
0.444
0.910
0.628
0.928
0.127
0.265
0.174
0.341
0.893
0.546
0.891
0.142
0.410
0.101
0.542
1.106
0.617
0.924
0.171
0.307
0.163
0.469
1.050
0.647
0.936
0.194
0.202
0.237
0.292
0.877
0.483
0.856
0.188
0.302
0.169
0.456
1.027
0.653
0.938
0.199
0.298
0.225
0.432
0.921
0.526
0.881
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Leaf Morphol
ogical Featu
r
e Extra
c
tion o
f
Digital Im
ag
e Anthoce
p
h
a
lus… (Fu
z
y Yustika M
ani
k)
635
(a)
(b)
Figure 7. Cal
c
ulatio
n morp
hologi
cal feat
ure
s
(a
) Leaf
spot di
sea
s
e
(b) L
eaf blight
disea
s
e
In experim
ent
s u
s
ing 1
00 i
m
age
s of ea
ch cla
s
s Ja
bo
n symptom
a
tic leaf
seedli
n
gs
were
tested to determin
e
whi
c
h
features a
r
e
cap
able to
re
pre
s
ent the image so that to be able to get
useful info
rm
ation and ca
n be use
d
to get good
accurate re
sults duri
ng the cla
ssifi
cat
i
on
pro
c
e
ss. Matrix resulted fro
m
the extraction of
morph
o
l
ogical traits the entire ima
ge in the form
of a m
a
trix m
easurin
g 7
x
100
whi
c
h
is
a repres
entat
ion of
100
im
age
s fo
r e
a
ch
type of
disea
s
e
with every im
age ha
s a ve
ctor
which is
comp
osed of 7 element
s.
Features
su
ch as area, pe
rimeter, m
a
jo
r ax
is
and
mi
nor
axis a
s
d
i
scusse
d can
not b
e
use
d
inde
pen
dently as o
b
j
e
ct ide
n
tificat
i
on featur
es.
Such featu
r
e
is influen
ce
d
by the size
of
the obje
c
t. In
ord
e
r
not to
dep
end
on
scaling,
som
e
of the fe
ature
s
that
ca
n
be d
e
rive
d f
r
om
these
featu
r
e
s
a
r
e
recta
n
gularity,
com
pactn
ess,
el
o
ngation, ecce
ntricity
, convi
c
ity, rou
ndn
e
ss
and solidity.
Rou
ndn
ess a
nd re
ctang
ul
arity sho
w
s h
o
w well
a
n
o
b
ject can be
descri
bed by
a circl
e
and a re
ctan
gle. While th
e comp
actn
e
ss me
asur
es
the ratio between the obje
c
t area and
circle
area u
s
in
g pe
rimeter. Ba
se
d on the Tabl
e 2 above, blight has a rou
ndne
ss value
(<0.20
7) an
d
a
recta
ngul
arity
(<0.52
3),
while le
af spo
t
has
a
ro
u
ndne
ss hi
gh
est
re
solutio
n
(>0.5
69) a
nd
recta
ngul
arity
(>0.73
8). Th
e averag
e value compa
c
t
ness leaf sp
ot is large
r
(>0.73
) than the
averag
e valu
e of blight (<0.237). Elon
g
a
tion sh
ow
s
elong
ated pol
ygon level. It
is kn
own as l
a
te
blight elo
nga
tion value
s
g
r
eate
r
(>
0.4
83)
of
the v
a
lue of
elon
gation at
l
e
a
f
spot
(<0.00
7).
Ecce
ntri
city is the
ratio
be
tween
the m
a
jor
axis
and
a min
o
r axis, from the
ta
ble a
bove
sh
o
w
s
that the val
u
e
of e
c
centri
cit
y
bli
ght
(<0.8
56) is g
r
eate
r
than
the va
l
ue of
e
c
centri
city leaf
sp
ot
(>
0.585).
Conv
exity and soli
dity are abl
e to de
scri
be
co
nvex of a pol
ygon. The diff
eren
ce
betwe
en
these t
w
o m
e
trics is t
hat
conve
c
ity usi
ng the ra
tio
of the peri
m
e
t
er whil
e the
solidity u
s
e a
r
ea
ratios. If the convexity calculate the relat
i
ve am
ounts t
hat differ fro
m
the convex
hull obje
c
t while
the obje
c
t de
nsity co
untin
g soli
dity.Lea
f spot ha
s th
e high
est
con
v
exity value (>1.18
4) a
nd t
he
highe
st soli
dity (>0.91
7) when compa
r
e
d
with the
value of convexi
t
y blight (<1.1
06) an
d soli
di
ty
leaf blight (>0
.
731).
(a)
(b)
Figure 8 Co
n
v
exhull perim
eter (a
) Leaf
blight (b
) Leaf
spots
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 14, No. 2, June 20
16 : 630 – 63
7
636
As
see
n
in
Fi
gure
8,
a
co
n
v
ex polygon
t
hat
seem
s
li
ke having a co
mplex detail that the
perim
eter
co
uld b
e
h
uge
comp
ared to
the p
e
rim
e
ter
of the
co
n
v
exhull, this
seem
s to b
e
the
symptom
s
of
a blig
ht of le
aves. O
ne
po
int at the
con
v
exity is a
co
nvex hull
(8a), if it is great
er
than on
e p
o
in
t the obje
c
t is not convex full or i
r
regul
a
r
bo
und
arie
s.
Base
d on
th
e table
above
2,
it is
kn
own th
at the
co
nvex leaf
sp
ot is
not convex fu
ll or ir
reg
u
lar
boun
dary
obj
ects,
whil
e l
a
te
blight h
a
s
a v
a
lue
app
roa
c
hing
co
nvexhull. Whil
e the
re
sult of
soli
dity is
kno
w
n
that the
den
sity
of object le
a
f
spot is hi
g
her tha
n
the
den
sity of obje
c
t leaf b
light. Solidity value obtai
ned
according
to the sym
p
tom t
hat occu
rs a
s
both
symp
to
ms of le
af bli
ght and
leaf
spot.
When
the
f
o
rm be
co
me
s le
ss
su
bt
le
(or
ro
ugh
) o
r
ha
s int
r
i
c
at
e detail
s
that
perim
eter
ca
n be ve
ry large
comp
ared to
its convex
hu
ll pe
rimeter,
so the
va
lue
could
be l
o
wer co
nvexity wh
ile the
value
of
high solidity indicates that
solidity is bet
t
e
r whil
e convexity is more sensitive.
(a)
(b)
Figure 9. Morpholo
g
ical fe
ature
s
jabo
n leaf
dise
ase symptoms (a)
Leaf sp
ots (b
) Blight
From the results of Figure
9, it is known t
hat the two Jabo
n leaf disea
s
e
symptoms an
d
see
n
that the value of the conv
exity is greater tha
n
the solid
ity
.
Leaf spot has the hi
gh
est
convexity an
d solidity value
comp
ared
to the val
u
e
of convexity and
solidity leaf blight. If
it is
see
n
from th
e value of el
o
ngation
blight
and le
af sp
ot, it is also
kn
o
w
n that the
symptoms of l
eaf
blight have a
n
elong
ated shape. It is kn
own that l
eaf
blight elon
gat
ion value is
g
r
eate
r
than th
e
value of elo
n
gation at le
af spot. Elong
a
t
ion
sh
owed
elong
ated pol
ygon level. If viewed from t
he
eccentri
city blight and leaf
spot, it is a
l
so kn
own that the symptoms of
leaf blight have a
n
elong
ated sh
ape. It is kn
own a
s
late
blight ecce
ntricity value g
r
eate
r
than t
he value of the
eccentri
city leaf sp
ot. Le
af spot h
a
s
a ro
undn
ess value an
d the hig
h
e
s
t rectan
gula
r
ity. This
according
to
the symptom
s
seen
that l
eaf sp
ot
ha
s
a sh
ape
like
a ci
rcl
e
. The
value of the l
eaf
spot
s for com
pactn
ess val
ue is greate
r
than the av
erage value of
blight. This in
dicate
s that the
leaf sp
ots h
a
v
e a more compa
c
t wh
en
comp
ared
with leaf blight
. This is
co
n
s
iste
nt with t
h
e
visible sympt
o
ms.
4. Conclusio
n
Utilizin
g the
sha
pe
as a
comp
one
nt i
n
the
im
age
analy
s
is to
investigate
th
e shap
e
feature
s
ca
n
be used to measure the characte
ri
sti
cs of the sh
ape of the object imag
e. To
measure the
basi
c
g
eom
etric attri
bute
s
, there ar
e
seven m
o
rph
o
logi
cal featu
r
es sel
e
cte
d
for
analysi
s
. Seven of the morpholo
g
ical fe
ature
s
nam
el
y convexity, solidity, elonga
tion, round
ne
ss,
recta
ngul
arity
,
eccentri
city, and
compa
c
tness. Each
i
s
u
s
e
d
to m
e
asu
r
e th
e
co
nvexity, solidi
t
y,
elong
ation, round
ne
ss, rectan
gle, ellipse, and
m
easure the
compl
e
xity of the form.
The
morp
holo
g
ica
l
feature
s
we
re abl
e to d
e
scrib
e
the
chara
c
te
risti
c
s of the sh
ap
e of the
sa
me
asp
e
ct. The
s
e feature
s
a
r
e very goo
d a
nd app
ro
priat
e
for u
s
e in
chara
c
te
rizi
ng
cla
s
ses of
Ja
bon
le
a
f
d
i
s
e
a
s
e
symp
t
o
m
s
ie
, le
a
f
s
p
o
t
a
n
d
le
a
f
b
lig
h
t. T
h
e
r
e
su
lts
s
h
ow
e
d
th
a
t
a
ll symp
t
o
m
s
sh
ape
prop
ertie
s
can be q
u
a
n
titatively explaine
d
by the feature
s
of shap
e. Overall
se
ven
morp
holo
g
ica
l
feature
s
are cap
able to
extract
characteri
stic
sy
mptom form
s contain
ed in
the
leaves
Jab
o
n
gets u
s
efu
l
information
from
an im
age. Leaf spot has a v
a
lue ro
und
n
ees,
comp
acte
ne
s and
recta
n
g
u
larity is g
r
ea
ter tha
n
the
b
light. The
s
e
result
s a
r
e
co
nsi
s
tent
with
the
symptom
s
se
en in fisu
al th
at leaf spot
symptoms h
a
ve a more co
mplex sh
ape,
while late bli
ght
symptom
s
h
a
ve elongate
d
sha
pe, it is prove
d
by the value of elongation
and leaf blig
ht
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Leaf Morphol
ogical Featu
r
e Extra
c
tion o
f
Digital Im
ag
e Anthoce
p
h
a
lus… (Fu
z
y Yustika M
ani
k)
637
eccentri
city greate
r
tha
n
leaf spot.
While th
e solidity and convexity able to de
scrib
e
the
symptom
s
of a more
conve
x
leaf spots a
nd soli
d from
the blight.
Referen
ces
[1]
W
a
h
y
ud
i. An
al
isis P
e
rtumbu
han
Da
n
Ha
si
l T
anaman
Ja
bon
(Anth
o
ce
pha
llus
C
ada
mba).
Jur
n
al
Perennial
. 201
2; 8(1): 19-24.
[2]
Prana
nd
a R, Indri
y
a
n
to, Ri
ni
arti M. Respo
n
Pertumb
uha
n Bibit Ja
bo
n
(Anthoce
p
h
a
lu
s Cada
mba
)
den
ga
n Resp
on Pertumb
u
han Bi
bit Ja
bon Pem
beri
a
n Kompos K
o
toran Sa
pi
Pada Me
dia
Pe
ny
ap
i
h
a
n
.
J
u
rna
l
Sylva Le
stari
. 2014; 2(
3
)
: 29-38.
[3]
W
i
d
y
astuti SM
, Harjon
o, Sur
y
a Z
A
. Infeksi
A
w
a
l
Jam
u
r
Urom
ycla
di
um tepper
ian
u
m
pad
a Da
un
F
a
lcatari
a
mol
u
ccan
a
da
n Acacia ma
ngi
um di La
borat
oriu
m.
Jurnal Man
a
je
men Huta
n
T
r
opika
. 20
13.
[4]
Anggr
ae
ni I, L
e
la
na NE. P
e
n
y
ak
it Karat T
u
mor Pad
a
Se
n
gon. Ba
da
n Pe
neliti
an
da
n Pe
ngem
ban
ga
n
Kehuta
n
a
n
. Jakarta. 2011.
[5]
Anggr
ae
ni, W
i
bo
w
o
.
Pen
g
e
n
dali
a
n
C
y
l
i
ndr
o
c
ladi
um S
p
. Pe
n
y
eb
ab P
e
n
y
a
k
it Lo
doh
Pa
da
Bibit
Acaci
a
Mang
ium W
ild
. Denga
n F
u
n
g
i Anta
gon
is T
r
ichoderma S
p
. dan Gli
o
cla
d
ium Sp.
Ju
rna
l
Pe
ne
l
i
t
ian
H
u
ta
n
Ta
nam
an
, Pu
sa
t L
i
tb
ang
H
u
ta
n
Ta
nam
an
. 200
9; 6(4
)
.
[6]
Herli
y
a
n
a
EN,
Achmad, P
u
tra
A. Peng
aruh
pup
uk or
gan
ik
cair terh
ad
ap
pertumb
uh
an
bibit J
a
b
on
(
Anthoce
p
h
a
lu
s cada
mb
a
mi
q.) dan ket
a
h
ana
nn
ya ter
h
a
dap
pen
ya
kit.
Jurna
l
Silvik
ult
u
r Tropika
.
201
2; 3(3): 168
-173.
[7]
Aisah
AR. Kl
as
ifikasi
da
n Pat
oge
nisitas
Ce
n
d
a
w
a
n
P
e
n
y
e
b
ab Prim
er P
e
n
y
ak
it Mati P
u
c
u
k p
ada
Bib
i
t
Jabo
n (
Anthoc
eph
alus C
a
d
a
m
b
a
(Ro
x
b.) Miq). T
e
sis. Bogor: Institut Pertani
an Bo
gor; 2
014.
[8]
Yunasfi. F
a
ktor
-F
aktor
y
a
ng M
e
mpe
ngar
uh
i Perkemb
ang
an
Pen
y
ak
it dan P
e
n
y
ak
it
y
a
ng D
i
seb
abka
n
ole
h
Jamur. Meda
n: USU Dig
ital Li
brar
y; 20
02.
[9]
Agrios GN. Pla
n
t Patholo
g
y
. F
i
fth editio
n
. Ne
w
York (US): E
l
sevier Ac
ade
mic Pr. 2005.
[10]
Gue L, Riven
o
D, Derado J,
Muntean
u C
R
, Pazos A. Automatic featur
e extr
action u
s
ing g
e
n
e
tic
progr
ammin
g
:
An a
ppl
icatio
n
to epi
le
ptic EE
G classificati
on
.
Expert Syste
m
s w
i
th A
ppl
ic
ations
. 201
1;
38: 104
25
–1
04
36
[11]
Ahsan
M, M
Dzu
l
kifli. Fe
at
ures E
x
tract
i
on
for Ob
jec
t
Dete
ction B
a
sed on Interest Point.
T
E
LKOMNIKA T
e
leco
mmunic
a
tion, Co
m
puti
ng, Electron
ics
and Co
ntrol.
2
013; 11: 2
716-
272
2.
[12]
Gonzal
ez R, W
oods, R. Dig
it
a
l
Image Proces
sing. T
h
ird edit
i
on.
Ne
w
J
e
rse
y
, USA: Pears
on Prentic
e
Hall. 2
008.
[13]
Putzu L, Ca
oc
ci G, Di Ru
be
rt
o C. Leuc
oc
yte cl
assific
a
ti
on for l
euk
ae
mia d
e
tection
usin
g ima
ge
process
i
ng tec
hni
ques. 2
014.
[14]
Z
i
novev
D, R
a
i
c
u D, F
u
rst J,
Armato SG. Predi
cti
ng
Rad
i
o
l
ogic
a
l P
a
n
e
l O
p
ini
ons
Usi
ng
a Pa
nel
of
Mechi
ne Le
arn
i
ng Cl
assifi
ers.
Algorit
h
m
s
. 20
09; 2: 147
3-15
00.
[15]
Gartner A, Li
n
nema
nn U, S
a
ga
w
e
A, Hofm
ann M, U
llric
h
B, Kleber
A. Morph
o
lo
g
y
of
zircon crsta
l
grai
ns i
n
se
dim
ents- ch
aracter
i
st
ics, classific
a
tions,
defin
itio
ns.
Jour
nal
of central Euro
pe
an
Ge
olo
g
y
.
201
3; 59: 65-7
3
.
[16] Anggr
ein
i
I.
Collet
o
trichu
m S
p
., Pen
y
eb
ab
Pen
y
ak
it Bercak Dau
n
Pad
a
Bebera
pa Bi
b
i
t
T
anaman
Hutan D
i
Perse
m
aia
n
. Pusat L
i
tban
g Huta
n T
anam
an. 20
11.
[17]
G Patil B, M
ane
N, S
ubb
ar
ama
n
S. Iri
s
F
eature
E
x
t
r
action
an
d
C
l
assificati
on
u
s
ing
F
P
GA.
Internatio
na
l Journ
a
l of Electr
ical
a
nd Co
mp
uter Engi
ne
erin
g (IJECE).
201
2: 214-2
22.
[18]
F
anan
i A, Yu
ni
arti A, Suci
at
i
N. Geometric
F
eature E
x
trac
tion of
B
a
tik I
m
age
Usin
g C
a
rdi
nal
Spl
i
n
e
Curve Re
pres
entatio
n.
TELKOMNIKA Teleco
mmu
n
icati
o
n, Com
puti
ng,
Electronics a
nd Co
ntrol.
201
4.
Evaluation Warning : The document was created with Spire.PDF for Python.