Int
ern
at
i
onal
Journ
al of Inf
orm
at
ic
s
and
Co
m
munic
at
i
on
Tec
hn
olog
y (IJ
-
I
CT)
Vo
l.
6
,
No.
3
,
D
ece
m
ber
201
7
, pp.
209
~
2
17
IS
S
N:
22
52
-
8776
,
DOI: 10
.11
591/iji
ct
.
v6
i
3.p
p
20
9
-
2
17
209
Journ
al h
om
e
page
:
http:
//
ia
esj
ou
r
nal.co
m/
on
li
ne/in
dex
.php
/
IJ
ICT
Density
Based Cl
usterin
g with Int
egrated
On
e
-
Class SVM
for
Noise Re
du
ctio
n
Md. A
bdul
A
w
al
,
M
oham
mad
Jah
angir
A
la
m
*
,
Md.
Nu
rul
Must
afa
Depa
rtment
o
f
C
om
pute
r
Scie
n
ce a
nd
In
form
at
ion
Technol
og
y
,
So
uthe
rn
Univ
ersity
B
anglade
sh
739/A
Mehidi
b
a
g
Road, Chitta
go
ng,
Bang
la
d
esh.
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
21
st
,
20
1
7
Re
vised
Oct
5
th
, 201
7
Accepte
d
Oct
19
th
,
20
1
7
The
t
ec
hno
log
y
base
d
m
oder
n
a
gric
ult
u
re
indust
rie
s
ar
e
tod
a
y
‟s
re
quire
m
ent
in
eve
r
y
p
art
of
agr
ic
u
lt
ure
in
B
angl
ad
esh.
In
this
te
chnol
og
y
,
th
e
disea
se
of
pla
nts
is
pr
ec
is
ely
cont
rol
le
d
.
Du
e
to
the
v
ariabl
e
at
m
ospheric
ci
r
cums
ta
nce
s
the
se
cond
it
ions
som
et
im
es
the
fa
rm
er
doesn‟t
k
now
what
t
y
p
e
o
f
disea
se
on
the
pla
n
t
and
w
hic
h
t
y
pe
o
f
m
edi
ci
n
e
provide
t
hem
to
avoi
d
di
sea
ses.
Thi
s
re
sea
rc
h
d
eve
lo
ped
for
cro
ps
disea
ses
det
ection
and
to
provide
s
soluti
on
b
y
using
image
proc
essing
te
chn
iqu
es.
W
e
have
use
d
Android
Studio
to
deve
lop
the
s
y
s
te
m
.
The
cro
ps
disea
ses
d
et
e
ct
ion
and
solu
ti
on
s
y
s
te
m
is
co
m
par
ed
the
image
of
aff
ecte
d
cro
ps
with
dat
aba
se
of
CDD
AS
S
(Cro
ps
Disea
ses
Dete
c
ti
on
and
Soluti
on
s
y
stem
).
If
CDD
AS
S
de
te
c
t
an
y
d
isea
se
s
y
m
ptom
,
the
n
provide
su
ggesti
on
so
tha
t
fa
rm
ers
ca
n
ta
ke
prope
r
dec
ision
to
provide
m
edi
ci
ne
to
the
aff
e
cted
cro
ps
.
The
applic
at
io
n
has
dev
el
op
e
d
with
user
frie
ndl
y
fe
at
ure
s
so t
ha
t
f
armers
ca
n
use
i
t
e
asily
.
Ke
yw
or
d:
agr
i
c
ultur
al
e
xperts
a
ndr
oid
a
pps
CDD
AS
S
Crops
Diseases
D
et
ect
ion
Im
age p
r
ocessi
ng
Copyright
©
201
7
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Moh
am
m
ad
Jahangir
A
la
m
,
Dep
a
rtm
ent o
f C
om
pu
te
r
Scie
nce a
nd Info
rm
at
ion
Tec
hnol
ogy, S
outhe
rn Unive
rsity
Bang
l
ades
h
739/A M
ehi
diba
g
R
oad, Chitt
agon
g,
Ba
ngla
des
h
.
E
m
a
il
:
j
ahangir@s
outher
n.
e
du.
bd
1.
INTROD
U
CTION
Agricult
ure
a
nd
hum
an
so
ci
al
dev
el
op
m
ent
go
si
de
by
s
ide
as
the
pro
du
ct
io
n
of
cr
ops
m
ade
it
po
s
sible
for
pr
i
m
itive
m
an
to
set
tl
e
down
in
sel
ect
ed
s
po
ts
l
eadin
g
t
o
f
orm
at
ion
of
s
ociet
y
[1
]
.
The
hist
or
y
of
agr
ic
ultur
al
in
Ba
ng
la
des
h
long
befor
e
.
B
ang
la
des
h
is
ba
sic
al
ly
an
agri
cultural
co
untry
,
an
d
the
i
nc
om
e
is
base
d on the a
gr
ic
ultur
al
pro
du
ct
s
and all
it
s r
es
ources
d
e
pe
nd on t
he
a
gr
i
cultural
outp
ut.
Althou
gh
Ba
ngla
des
h
is
on
course
f
or
Mi
d
dle
I
nc
om
e
Cou
nt
ry
sta
tus
by
2021,
ag
ricult
ur
e
rem
ai
ns
the largest
em
plo
ye
r
in the c
ount
ry b
y fa
r
an
d
47.
5% of
t
he
p
op
ulati
on
is
directl
y e
m
plo
ye
d
in agric
ult
ur
e a
nd
arou
nd
70%
de
pends
on
a
gr
i
culture
in
one
form
or
an
othe
r
f
or
their
li
velihood.
Agricul
ture
is
t
he
s
our
ce
of
foo
d
for
pe
op
l
e
through
cr
op
s,
li
vestock,
fisher
ie
s;
the
source
of
ra
w
m
at
erial
s
fo
r
in
dustry,
of
tim
ber
for
const
ru
ct
io
n;
and
a
gen
e
rat
or
of
f
orei
gn
exc
hange
f
or
the
c
ountry
thr
ough
the
e
xport
of
a
gr
ic
ultur
al
com
m
od
it
ie
s,
wh
et
her
ra
w
or
processe
d.
It
is
the
m
oto
r
of
the
de
velo
pm
ent
of
the
agro
-
i
ndus
tria
l
sect
or
includi
ng
f
ood
processi
ng,
in
put
pro
duct
io
n
a
nd
m
ark
et
ing
,
an
d
relat
ed
ser
vices. A
s
m
ai
n
source
of
ec
onom
ic
li
nk
age
s
in
r
ural
areas,
it
play
s
a
f
undam
ental
ro
le
in
re
duci
ng
pove
rt
y
w
hich
rem
ai
ns
a
pr
e
dom
inantly
rura
l
ph
e
nom
eno
n
[
2].
Th
ough
,
i
n
A
gri
cultural
De
pa
rtm
ent
te
chn
ol
og
y
is
ra
pid
ly
changin
g,
m
any
autom
at
ic
tech
no
lo
gies
are
com
ing
in
the
m
ark
et
(e
xam
ple
,
Au
to
m
at
ic
planting,
cutte
r
m
achines
et
c
wh
ic
h
helps
the
farm
er
to
pro
du
ce
m
axim
u
m
pr
oducts
).
Plant
diseas
e
is
an
i
m
po
rtant
con
c
er
n
f
or
the
far
m
ers
in
Ba
ngla
des
h
because
Plant
disease
i
s
an
im
pair
m
e
nt
of
t
he
norm
al
sta
te
of
the
plant
t
hat
inter
rupts
or
m
od
if
ie
s
it
s
vital
f
unct
ions
[3]
.
To
get
s
olu
ti
on
on
pla
net
disease
I
f
farm
ers
deci
de
t
o
ta
ke
a
dvic
e
from
agr
ic
ultu
ra
l
ex
per
t
regar
di
ng
the
treatm
ent
of
in
ci
den
ce
of
pest
/disea
se/
trai
t
to
their
cr
op
/
plant
in
order
t
o
increase
t
he
cr
op
pro
duct
ivit
y
the
n
he
m
ay
f
ace f
ol
lowing situat
i
on
s
[4]:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2252
-
8776
IJECE
V
ol.
6
,
No.
3
,
Decem
ber
20
1
7
:
209
–
2
17
210
i) S
om
eti
m
es t
hey h
a
ve
t
o go
lon
g dist
ances
for
a
ppr
oach
i
ng the
expe
rt.
ii
)
Eve
n
th
ou
gh they
go s
uc
h dist
ances e
xper
t
m
ay
n
ot be a
vaila
ble at t
hat
tim
e.
ii
i)
So
m
eti
m
es
,
the
ex
pe
rt
w
ho
m
a
far
m
er
con
ta
ct
s,
m
ay
no
t
be
in
a
posit
ion
to
a
dv
is
e
the
far
m
er
with the
av
ai
la
ble in
form
at
io
n
a
nd k
nowle
dge.
In
these
cases
seekin
g
the
ex
per
t
ad
vice
is
ver
y
ex
pe
ns
iv
e
and
tim
e
con
su
m
ing
.
He
nc
e
el
ect
ro
nic
exp
e
rt
syst
em
s
are
nee
ded.
El
ect
ronic
ex
pe
rt
syst
e
m
s
enab
l
e
far
m
ers
in
i
de
ntifyi
ng
ty
pe
of
diseases;
m
akin
g
the
ri
gh
t
decisi
on
an
d
s
el
ect
ing
the
pro
per
treat
m
ent.
The
exp
e
rt
syst
em
s
are
i
ntell
igent
com
pu
te
r
pro
gr
am
s
that
are
capa
ble
of
offe
rin
g
s
olu
ti
ons
or
ad
vi
ces
relat
ed
to
sp
eci
fic
pro
ble
m
s
in
giv
en
dom
ai
n,
bo
t
h
in
a
way
and
at
a
le
vel
com
par
able
to
that
of
hu
m
an
exp
e
rt
in
a
fiel
d.
On
e
of
the
a
dvan
ta
ges
of
us
i
ng
Ele
ct
r
onic
exp
e
rt
syst
e
m
s
is
it
s
a
bili
ty
to
reduc
e
the
inform
at
i
on
that
hum
an
us
ers
nee
d
to
proces
s,
re
du
ce
per
s
onnel
co
sts
and
increase
t
hroughp
ut.
A
nothe
r
ad
va
ntage
of
ex
per
t
syst
em
is
that
it
perf
or
m
s
ta
sk
s
m
or
e
co
ns
ist
ently
tha
n
hu
m
an
ex
pe
rts
[5
]
.
Ther
e
a
re
var
i
ou
s
kind
s
of
a
bnorm
ality
s
ta
t
es
pr
es
ent
on
t
he
pla
nts
le
af
wh
ic
h
can
be
i
den
ti
fie
d
by
m
ean
of
m
anu
al
insp
ect
ion.
The
i
m
age
pro
cessi
ng
a
nd
pa
tt
ern
recog
niti
on
te
ch
niques
play
the
wo
rt
h
fu
ll
ro
le
to
c
onver
t
m
a
nu
al
proce
ss
t
o
a
uto
m
at
e
the
proces
s.
The
autom
at
ic
diagn
o
sis
syst
em
based
on
plant
disease
featur
e
s
re
du
ce
s
the
de
pende
nc
y
on
e
xp
e
rts
in
the
area
c
on
cern
e
d
[6
-
8].
D
ependin
g
on
t
he
app
li
cat
ion,
m
any
of th
os
e
pro
blem
s
m
ay
be
so
l
ved or at l
east
reduce
d
by the
us
e
of im
age p
ro
ces
sin
g.
Hen
ce
we
are
proposing
an
autom
at
ic
crops
diseases
de
te
ct
ion
an
d
sol
ution
syst
em
(CD
DAS
S
)
wh
ic
h
help
s
a
far
m
er
to
identify
the
disease
of
cr
ops
an
d
pr
ov
i
de
treat
m
en
t
fo
r
his
cr
ops
accor
ding
to
di
sease
us
in
g
i
m
age
processin
g
te
ch
ni
qu
es
with
out
help
of
any
cr
op
s
diseases
exp
e
rt.
To
get
s
olu
ti
on,
fa
rm
e
rs
just
n
eed
to
instal
l
the
a
pps
i
nto
their
m
ob
il
e
phone
s
f
or
the
f
irst
tim
e
and
uploa
d
t
he
a
ff
e
ct
ed
cr
ops
‟
im
age
t
o
detect
an
d
get
su
ggest
io
n
fro
m
the
syst
e
m
.
That‟s
way
a
f
arm
er
can
save
tim
e,
eff
ort
s
a
nd
m
on
ey
a
nd
can
be
te
ns
io
n
f
ree.
2.
LIT
ERATUR
E REVIE
W
Pr
of.
H.
M.
Des
hm
uk
h,
Ja
dha
v
Sanjiva
ni,L
ohar
Ut
karsha
,
Bhagat
Ma
dhuri
and
Sal
unke
Sh
ub
hangi
.
“Pla
nt
Lea
f
D
ise
ase
I
den
ti
fi
cat
ion
Syst
em
for
A
ndro
i
d”.
In
te
rn
at
io
nal
Jo
ur
nal
of
A
dvance
d
Re
sea
r
ch
i
n
Com
pu
te
r
an
d
Com
m
un
ic
at
i
on
En
gin
ee
rin
g
[
9
]
.
I
n
this
arti
cl
e
they
ha
ve
de
scri
bed
t
he
de
vel
op
m
ent
of
an
Androi
d
a
pp
li
c
at
ion
th
at
gi
ve
s
us
e
rs
or
fa
rm
ers
the
ca
pab
il
it
y
to
identify
t
he
plant
le
af
di
seases
based
on
the
photog
raphs o
f
p
la
nt lea
ves
ta
ken th
rou
gh an an
droid
appli
cat
ion
.
Su
fiy
a
n
K
S
he
kh, A
nik
et
Ba
it
ule, Mil
ind
N
arethe,
Sa
ngap
pa
Ma
ll
a
d,
W
a
ghda
rikar
a
nd D
r D Y
Pati
l
.
“Detec
ti
on
of
Leaf
Diseases
and
Mo
nito
rin
g
the
A
gr
ic
ultur
al
Re
source
s
us
ing
And
ro
i
d
A
pp
”
.
In
te
rnat
ion
al
Jo
ur
nal
of
I
nnov
at
ive
Re
se
ar
ch
in
C
om
pu
te
r
an
d
Com
m
un
ic
at
ion
En
gin
e
erin
g
[
10
]
.
In
t
his
arti
cl
e
they
ha
ve
descr
i
bed
a
nd
desig
ne
d
a
ne
w
arc
hitec
ture
for
rem
ote
co
ntr
ol
of
a
gri
cu
lt
ur
e
de
vices
a
nd
al
s
o
de
te
ct
ing
t
he
diseases
of p
la
nt which
m
ake m
uch
easie
r
a
nd less
de
pende
nt of t
he
c
onditi
on
s
prese
nt to farm
ers.
Mr.S
us
ha
nt.S
.
Chava
n,
Mr.
Nite
sh
.
P.S
at
re
,
Mr.Ra
j
at
.R.
D
eshm
ane,
Mr.
Suhas.B.
Katka
r
.
„„A
ndro
i
d
Ba
sed
App
t
o
Pr
e
ven
t
C
rop
Diseased
in
V
ario
us
Sea
sons
‟‟
.
I
nter
nationa
l
Re
search
Jou
rn
al
of
E
ng
i
ne
erin
g
and
Tec
hnol
og
y(IRJET
)
[
11
]
.
In
this
a
rtic
le
they
ha
ve
de
velop
e
d
a
n
a
ndr
oid
a
pp
li
c
at
ion
f
or
a
gri
cultu
re,
whe
n
and whic
h fe
rtil
iz
ers,
p
e
sti
ci
des
an
d her
bicid
es are
us
e
d
t
o
s
ave
var
i
ou
s
cr
ops
from
v
ario
us disease
s
.
3.
PROP
OSE
D MET
HO
D
3.1
.
Flow
c
hart
Figure
1.
Flo
w
char
t
of Cr
ops
Diseases
Detec
ti
on
a
nd S
olu
ti
on Syste
m
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
ICT
IS
S
N:
22
52
-
8776
Crop
s
D
ise
as
e
s D
et
ect
io
n and Sol
ution Sy
ste
m
(
M
d. A
bdul
Awal
)
211
The
flo
wc
har
t
descr
i
bes
the
whole
process
of
the
resear
c
h.
At
first
us
e
r
has
to
uploa
d
a
def
ect
e
d
crops
im
age o
n t
he
syst
e
m
then
syst
e
m
w
il
l s
earch
im
age r
el
at
ed
inform
at
i
on
in
th
e syst
em
database,
if m
at
ch
the
up
l
oad
e
d
im
age
with
the
syst
e
m
database
it
wi
l
l
pr
ov
i
de
po
ssi
ble
diseases
nam
e
wit
h
so
luti
on
,
if
does
not
m
at
ch
it
w
il
l show n
ot m
at
ch
m
essage.
3.2
.
I
nt
er
f
ace
Design
Figure
2.
H
ome
p
a
ge of
Cr
ops D
ise
ases
D
et
ect
ion
a
nd S
ol
ution Syst
em
The
fig
ure
sho
ws
m
ai
n
pag
e
of
the
syst
em
wh
e
re
us
e
rs
ca
n
ch
oo
s
e
their
act
ivit
ie
s
including
m
enu
acce
ss and
dif
f
eren
t c
rops
li
st. A
us
e
r
ca
n be
able to
see s
ome
d
ise
ases
li
st wit
h
im
age.
Figure
3.
Disea
ses List
Pa
ge
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2252
-
8776
IJECE
V
ol.
6
,
No.
3
,
Decem
ber
20
1
7
:
209
–
2
17
212
Figure
4.
Disea
se w
ise
sym
ptom
an
d
so
l
ution Page
Af
te
r
cl
ic
king
any
ic
on
,
im
a
ge
of
disease
s,
visit
or
can
be
able
to
fin
d
s
pecific
disease
sy
m
pto
m
s
with
s
olu
ti
ons
.
4.
RESEA
R
CH MET
HO
D
Ther
e
are
so
m
e f
undam
ental
steps:
a.
Im
age
Acquisi
ti
on
:
Im
ages
of
the
infe
ct
ed
le
aves
hav
e
ta
ken
f
ro
m
the
crops.
T
he
database
ha
s
diff
e
re
nt ty
pes im
ages on plan
t diseases a
nd t
he
im
ages ar
e s
tore
d
in
(
.
jpg, .
png) im
age f
or
m
at
.
b.
Im
age
pr
ep
ro
c
essing:
I
f
nois
es
are
pr
ese
nt
in
i
m
age,
interest
ed
reg
i
on
in
the
i
m
age
is
no
t
cl
ear.
I
n
the
im
age
cl
ipp
in
g,
sm
oo
thi
ng,
e
nhancem
ent
is
the
three
st
eps
incl
ud
e
d
i
n
pr
e
proces
sin
g
ph
a
se.
T
he
process
of
im
age
colle
ct
io
n
and
lots
of
i
nfor
m
at
ion
m
a
y
br
in
g
no
ise
wh
ic
h
m
ake
the
qual
it
y
of
i
m
age d
r
oppe
d. T
o per
form
d
e
m
isi
ng
d
if
fe
re
nt
ki
nds
of
re
duct
ion t
ech
nique a
re a
pp
li
cab
le
.
c.
Im
age
Segm
en
ta
ti
on
:
Im
age
s
egm
entat
ion
is
the
first
ste
p
and
al
s
o
one
of
the
m
os
t
crit
i
cal
ta
sk
s
of
i
m
age an
al
ysi
s.
Acc
ordi
ng to t
he
re
gion
of in
te
rest, the
im
age w
il
l be
segm
ented
i
nto
dif
fe
ren
t
par
ts
.
d.
Im
age
Feat
ur
e
Extracti
on:
T
he
fe
at
ur
e
s
ext
racti
on
is
the
i
nput
data
tra
nsfo
rm
into
set
of
featu
res
.
The
featu
re
se
t
will
extract
the
releva
nt
inf
or
m
at
ion
.
so
sh
ould
care
f
ul
ly
hav
e
cho
s
en.
Feat
ure
extracti
on
in
volves
sim
plify
ing
t
he
am
ount
of
res
ources
re
qu
ire
d
t
o
de
scribe
a
la
r
ge
set
of
data
accuratel
y.
e.
Im
age
cl
assifi
c
at
ion
:
The
i
ntent
of
t
he
cl
assifi
cat
ion
proces
s
is
to
cat
egorize
al
l
pix
el
s
in
a
dig
it
a
l
i
m
age
into
on
e
of
cl
asses
or
t
hem
e.
The
ob
je
ct
ive
of
im
ag
e
cl
assifi
cat
ion
is
to
ide
ntify,
as
a
uniq
ue
gr
ay
le
vel
(
or
colo
r)
,
feat
ur
e
s
occ
urrin
g
i
n
an
im
age
in
te
rm
s
of
the
obje
ct
or
ty
pe
t
he
se
feat
ur
es
act
ually
rep
res
ent
on
the
gr
ound.
Im
age
cl
assifi
cat
ion
is
pe
rh
a
ps
the
m
os
t
i
m
po
rtant
pa
rt
of
di
gital
i
m
age an
al
ysi
s [
12
].
5.
BINAR
Y RO
BUST I
NVA
R
IAN
T
S
CA
L
A
BL
E KEY P
O
INTS
(BRI
SK
)
:
THE
METHOD
Descr
i
ption
of
the
key
sta
ges
in
BR
IS
K
,
na
m
el
y
featur
e
de
te
ct
ion
,
de
scri
ptor
com
po
sit
ion
a
nd
key
po
i
nt m
a
tc
hin
g t
o
the level o
f deta
il
that the m
ot
ivate
d
rea
de
r
can
unde
rsta
nd
a
nd r
e
pro
duce. I
t i
s i
m
po
rt
ant to
no
te
that
t
he
m
od
ularit
y
of
t
he
m
et
ho
d
al
lo
ws
the
us
e
of
t
he
BR
ISK
detect
or
in
c
om
bin
at
ion
with
a
ny
oth
er
key point
desc
r
iptor an
d vice
ver
sa
, optim
iz
i
ng for t
he desir
ed per
form
ance and the
task
at
h
an
d [
13
].
5.1.
Scale
Sp
ac
e Key P
oint
Det
ec
tion
W
it
h
the
ai
m
of
ac
hieving
in
var
ia
nce
to
sca
le
wh
ic
h
is
cr
uc
ia
l
fo
r
hi
gh
-
qual
it
y
key
po
in
ts,
we
go
a
ste
p
furthe
r
by
searching
f
or
m
axi
m
a
no
t
on
ly
in
the
i
m
a
ge
plane
,
but
al
so
in
scal
e
-
s
pace
us
in
g
the
FA
ST
scor
e
s
as
a
m
e
asur
e
f
or
sal
ie
ncy.
Des
pite
di
screti
zi
ng
t
he
scal
e
axis
at
c
oar
se
r
i
nter
vals
tha
n
in
al
te
r
nativ
e
high
-
perf
or
m
a
nce
detect
or
s
,
t
he
BR
ISK
det
ect
or
est
im
at
es
the
tr
ue
scal
e
of
eac
h
key
point
in
the
c
onti
nuous
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
ICT
IS
S
N:
22
52
-
8776
Crop
s
D
ise
as
e
s D
et
ect
io
n and Sol
ution Sy
ste
m
(
M
d. A
bdul
Awal
)
213
scal
e
-
sp
ace
.
I
n
the
BR
IS
K
f
r
a
m
ewo
r
k,
the
scal
e
-
sp
ace
pyram
id
layers
con
sist
of
n
oct
aves
c
i
an
d
n
intra
-
octaves
d
i
,
for
i = {0,
1, . .
., n
− 1
}
and ty
pic
al
ly
n
= 4.
The
octaves
a
re
f
or
m
ed
by
pro
gr
essi
vely
half
-
sam
pling
the
ori
gi
nal
i
m
age
(corr
es
pondin
g
t
o
c
0
)
.Each
int
ra
-
octave
di
is
locat
ed
in
-
be
twe
en
la
ye
rs
c
i
and
c
i+
1
(as
il
lustrate
d
in
Figure
5).
The
first
intra
-
octave
do
is
obta
ined
by
do
wn
sam
pling
the
or
igi
nal
im
age
c
0
by
a
fa
ct
or
of
1.5
,
w
hi
le
the
rest
of
t
he
intr
a
-
octave
la
ye
rs
are
der
i
ved
by
su
ccessi
ve
half
sam
pling
.
Therefo
re,
if
t
deno
te
s
scal
e
then
t
(c
i
)
=
2
i
an
d
t(
d
i
)
=
2
i
*1.
5 [
5].
Figure
5
.
Scal
e
-
sp
ace
interest
po
i
nt d
et
ect
io
n
Scal
e
-
sp
ace
i
nt
erest
point
dete
ct
ion
:
a
key
po
int
(i.e.
sal
ie
nc
y
m
axi
m
u
m
)
is
identifie
d
at
oc
ta
ve
c
i
by
analy
zi
ng
t
he
8
nei
ghbori
ng
sal
ie
ncy
sco
r
es
in
c
i
as
w
el
l
as
in
the
c
orres
pondin
g
s
cor
es
-
patches
in
the
i
m
m
ediat
e
ly
-
neighborin
g
la
ye
rs
ab
ove
an
d
belo
w.
I
n
al
l
three
la
ye
rs
of
i
nterest,
the
l
oc
al
sal
ie
ncy
m
a
xim
u
m
is
su
b
-
pi
xel
refi
ned
be
f
or
e
a
1D
par
a
bola
is
fitt
ed
al
on
g
t
he
scal
e
-
axis
to
determ
ine
the
t
ru
e
scal
e
of
th
e
key
po
i
nt.
The
lo
c
at
ion
of
the
ke
y
po
int
is
the
n
al
so
re
-
inter
po
la
te
d
betwe
en
the
patc
h
m
axi
m
a
cl
os
est
to
the
determ
ined
sca
le
[
13
].
5.2.
Key
P
oint Descri
p
tion
Give
n
a
set
of
key
points
(c
on
sist
in
g
of
s
ub
-
pix
el
re
fine
d
i
m
age
locat
ion
s
an
d
ass
ocia
te
d
floati
ng
-
po
i
nt
scal
e
val
ues),
the
BR
I
SK
descr
i
ptor
is
com
po
sed
a
s
a
bin
a
ry
string
by
co
ncate
nating
t
he
res
ults
of
si
m
ple
br
ig
htne
ss
com
par
iso
n
te
sts.
This
i
de
a
has
be
e
n
de
m
on
strat
ed
in
to
be
ver
y
e
ff
ic
ie
nt,
ho
wev
e
r
he
re
w
e
e
m
plo
y
it
in
a
far
m
or
e
qual
it
at
ive
m
ann
er.
In
BR
I
SK,
w
e
identify
the
char
act
e
risti
c
directi
on
of
ea
ch
key
po
i
nt
to
al
low
for
ori
entat
ion
-
norm
al
iz
ed
descr
ipto
rs
an
d
he
nce
achie
ve
r
otati
on
in
var
ia
nce
w
hich
is
ke
y
to
gen
e
ral
r
obust
ness.
Als
o,
w
e
caref
ully
sel
ect
the
bri
ght
ness
c
om
par
ison
s
with
the
f
ocus
on
m
axim
iz
ing
descr
i
ptivene
ss
[
13
].
6.
DESCRIPT
O
R MAT
CHIN
G
Ma
tc
hin
g
tw
o
BR
IS
K
de
scri
ptors
is
a
si
m
ple
com
pu
ta
ti
on
of
their
Ha
m
m
ing
distance
as
done
in
BR
IEF:
the
nu
m
ber
of
bits
di
ff
e
ren
t
in
the
t
wo
desc
riptors
is
a
m
easur
e
of
their
dissim
ilarity
.
No
ti
ce
th
at
the
resp
ect
ive
ope
rati
on
s
reduce
to
a
bitwise
X
OR
fo
ll
ow
e
d
by
a
bit
count,
wh
ic
h
can
both
be
com
pu
te
d
ve
ry
eff
ic
ie
ntly
on t
od
ay
‟s
a
rch
it
ec
tures [
13
]
.
7.
E
X
PERI
MEN
TAL RES
UL
TS A
ND DIS
CUSSIO
N
7.1.
Ex
peri
ment
al result
wi
th
ap
pl
ying al
go
ri
th
m
The
ex
pe
rim
e
nt
is
c
onduct
e
d
i
n
Android
Op
e
rati
ng
Syst
e
m
s
us
in
g
O
pe
nCv
31
0
wh
ic
h
us
e
d
im
age
processi
ng
a
nd
i
m
age
com
pari
so
ns
te
ch
niqu
es
with
BR
ISK
al
gorithm
.
Th
e
databa
se
is
cr
eat
ed
of
si
x
dif
fer
e
nt
plants
.
It
co
nsi
sts
of
30
im
a
ges
of
Pa
ddy
plant
af
fected
Ba
ct
erial
Bl
ig
ht
Disease,
B
act
erial
Leaf
Stre
a
k
Disease,
20
im
ages
of
Paddy
plants
af
fected
by
Fo
ot
Rot
D
ise
ase,
38
im
a
ges
of
Early
Bl
igh
t
an
d
Lat
e
Bl
igh
t
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2252
-
8776
IJECE
V
ol.
6
,
No.
3
,
Decem
ber
20
1
7
:
209
–
2
17
214
of
T
om
at
o
plant,
38
im
ages
of
Early
Bl
igh
t
an
d
Lat
e
B
li
gh
t
of
P
otato
plant,
20
im
a
ges
of
An
t
hr
a
cnose
Disease
of Mango
plant a
nd
15 im
ages o
f
R
ot
W
il
t Disea
s
e of J
ute p
la
nt
.
Table
1.
T
he
s
a
m
ple o
f
d
if
fe
r
ent
sta
ges
of a
f
fected
c
rops
store
d
i
nto
t
he d
at
abase
We
ha
ve
us
e
d
BR
IS
K
al
gorithm
in
ou
r
rese
arch
t
o
m
at
ch
the
uploa
de
d
a
ff
ect
ed
cr
ops
im
ages
with
our
database
im
ages.
As
pe
r
BR
ISK
al
gor
it
h
m
‟s
te
chn
i
ques
th
e
syst
em
ta
kes
100
ke
y
obj
ect
s
from
both
up
l
oad
e
d
im
age
e.g
.
cl
ass
1
a
nd
al
l
database
i
m
ages
e.g
.
cl
ass
2
for
the
s
pecific
cr
op
th
en
it
is
com
par
e
ke
y
obj
ect
s
be
twee
n
cl
ass
1
a
nd
c
la
ss
2.
A
fter
find
i
ng
t
he
m
at
c
hing
of
the
obje
ct
s
it
do
es
cal
culat
e
the
pe
rc
entage
of
m
at
ched
f
ou
nd. It
foll
ows t
he belo
w
e
qu
at
ion
:
If
cal
culat
e
d
pe
rcen
ta
ge
is
a
bove
70
the
n
di
sp
la
y
the
m
at
c
hed
im
age
fro
m
database
an
d
pro
vid
e
the
po
s
sible
diseas
e
nam
e
and
s
ol
ution
ab
out
th
e
disease
ot
herwise
prov
i
de
s
m
essage
“
Up
l
oad
e
d
im
age
is
not
m
at
che
d
wit
h
t
he
database”
.
7.2.
Resul
t
w
it
h d
e
velop
e
d
apps
Me
nu
b
a
r
is
a u
ser
inter
face w
it
h
diff
e
re
nt m
enu
it
em
l
ike
Das
hboa
rd,
T
ake
P
hoto
a
nd U
pl
oad
P
ho
t
o
et
c.
Figure
6.
Me
nu li
st
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
ICT
IS
S
N:
22
52
-
8776
Crop
s
D
ise
as
e
s D
et
ect
io
n and Sol
ution Sy
ste
m
(
M
d. A
bdul
Awal
)
215
Using
U
plo
a
d
Photo
m
enu
fa
rm
ers
can
up
l
oa
d
af
fected
cr
ops
ph
oto
to
de
te
ct
diseases
and
will
get
so
luti
on.
Figure
7.
U
ploa
d Im
age P
age
Af
te
r
cl
ic
king
Up
l
oad
P
ho
t
o
m
enu
it
e
m
far
m
ers
or
vis
it
or
s
ca
n
see
Photo
Up
l
oad
op
ti
on.
He
re
cl
ic
kin
g
us
e
rs
will
sel
ect
sp
e
ci
fic
crops,
us
e
r
can
sel
ect
ph
oto
by
ta
ppin
g/
cl
ic
kin
g
blan
k
i
m
age
and
will
cl
ic
k
Up
l
oad butt
on.
Figure
8.
U
ploa
d Im
age P
age
with
process
Di
al
og
Af
te
r
cli
ckin
g Up
l
oad butt
on
a proces
s
dialo
g wil
l sh
ow.
Figure
9.
S
ucc
essfu
l
d
et
ect
io
n resu
lt
pag
e
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2252
-
8776
IJECE
V
ol.
6
,
No.
3
,
Decem
ber
20
1
7
:
209
–
2
17
216
If
up
l
oad
e
d
photo
m
at
ched
w
it
h
store
d
im
a
ge
the
n
su
cce
s
sfu
l
detect
ion
resu
l
t
pa
ge
will
sh
ow
with
po
s
sible
diseas
e n
am
e, cau
ses
and s
olu
ti
ons.
Figure
10.
Fail
ed dete
ct
ion re
su
lt
p
a
ge
If
uploa
de
d
i
m
age
does
not
m
at
ch
with
stored
im
age
,
then
it
will
sh
ow
'Cou
ld
not
de
te
ct
disease'
m
essage.
7.3.
C
ompari
s
on
s
t
at
eme
nt
We
ha
ve
c
ompare
d
the
C
rop
s
Diseases
Detect
ion
an
d
S
ol
ution
syst
em
with
ot
her
s
sy
stem
and
got
the b
et
te
r
, e
ff
ic
ie
nt p
e
rfor
m
ance of th
e
syst
em
w
hich
sta
te
d i
n
the
b
el
ow t
able 2.
Table
2.
C
om
par
iso
n of Cr
ops D
ise
ases
D
et
ect
ion
a
nd S
ol
ution wit
h ot
he
rs
syst
em
s
Criteria
Cro
p
s
Diseas
es
Detectio
n
an
d
So
lu
tio
n
syste
m
Plan
t
Leaf
Diseas
e
Iden
tif
icatio
n
S
y
ste
m
f
o
r
An
d
roid
[
9
]
Detectio
n
o
f
Leaf
Diseas
es
an
d
Mon
ito
ring
th
e
Ag
ricultu
ral
Res
o
u
rces
u
sing
An
d
roid
Ap
p
[
1
0
]
An
d
roid
Bas
ed
Ap
p
to
Preven
t
Cro
p
Di
seas
ed
in
Variou
s Seaso
n
s
[
1
1
]
Eno
u
g
h
in
f
o
r
m
atio
n
sto
red in
d
atab
ase abo
u
t plan
t dis
eases
Yes
Yes
Yes
Yes
Fu
ll part of
the d
iseas
es
Yes
No
No
No
Many
k
in
d
s o
f
plan
ts
Yes
Yes
Yes
No
Easy
to u
se
Yes
Yes
Yes
Yes
Access f
ro
m
ev
er
y
where
Yes
Yes
Yes
Yes
A
ccurac
y
Yes
No
Yes
Yes
Faster to g
et
resu
lt
Yes
Yes
No
No
Detail
s
so
lu
tio
n
Yes
No
No
No
8.
ADV
AN
T
AGE
S
a.
It is easy
to
ins
ta
ll
b.
No n
ee
d
t
o reg
ist
er
c.
It is use
r frie
ndly
d.
It work
‟s
in
d
i
f
fer
e
nt v
e
rsion
of an
droid
e.
It is li
gh
t
wei
ght
f.
It is faste
r
9.
CONCL
US
I
O
N
This
Au
t
om
at
e
d
Cr
ops
Diseas
es
Detect
io
n
a
nd
S
olu
ti
on
Sy
stem
is
te
ste
d
in
diff
e
re
nt
A
ndr
oid
phone
and
it
is
wor
kin
g
nicel
y
and
t
he
syst
em
is
ver
y
us
ef
ul
f
or
f
arm
ers
especial
ly
wh
o
are
li
vi
ng
in
rural
are
a
and
wh
e
re
the
a
gri
cultural
e
xp
e
rt
s
are
not
a
vaila
ble
.
U
ser
ca
n
easi
ly
us
e
this
app
li
cat
io
n.
A
ny
pe
op
le
ca
n
detect
and
get
so
l
ution
a
ny
tim
e
us
ing
this
a
ppli
cat
ion
.
T
he
ap
plica
ti
on
is
fr
e
e
of
c
os
t
an
d
do
e
s
not
re
qu
i
re
any
add
it
io
nal
dev
i
ce.
Evaluation Warning : The document was created with Spire.PDF for Python.
IJ
-
ICT
IS
S
N:
22
52
-
8776
Crop
s
D
ise
as
e
s D
et
ect
io
n and Sol
ution Sy
ste
m
(
M
d. A
bdul
Awal
)
217
REFERE
NCE
S
[1]
A
y
esha
Bhandr
a
li
a
,
Resham
Ary
a,
S.
N.
Panda,
Sachi
n
Ahuja
.
“
Pol
y
house
Agric
ult
ura
l
Mark
et
in
g
Sy
stem
Us
ing
Big
Dat
a
Hadoo
p
.
”
In
te
rnat
ional J
ournal
o
f Adv
a
nce
s in
App
li
ed
Sci
en
ce
s (
IJ
AA
S
)
.
2016
;
5
(
2
):78
-
84.
[2]
Ham
id
Miah
.
“
Agric
ult
u
re
Se
ct
or
Deve
lopment
St
ra
te
g
y
:
ba
ckgr
ou
nd
pape
r
fo
r
pre
par
ation
of
7th
Five
Ye
ar
Plan
,
”
htt
p://ww
w.pl
an
comm
.
gov.
bd/7th
-
five
-
y
e
ar
-
pl
an/
[3]
Hite
shw
ari
Sabr
ol1,
Sati
sh
Kum
ar.
“
Rec
ogni
ti
on
of
Tomato
La
t
e
Bli
ght
b
y
using
DW
T
and
Com
p
onent
Anal
y
sis
,
”
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJ
ECE
)
.
2017;
7
(
1
):
194
-
199
.
[4]
MS
Prasad
Babu
and
B
Sriniv
asa
Rao
.
“
Le
av
es
Rec
ognition
Us
i
ng
Bac
k
Propag
at
ion
Neur
al
N
etw
ork
-
Advice
For
Pest
and
Dise
ase
Control On Cro
ps
,
”
Indi
aKisa
n.
Net:
Expe
rt
Adv
issor
y
S
y
s
te
m
.
2
007.
[5]
Ja
y
ama
la
K.
Pat
il
,
Ra
j
Kum
ar.
“
Plant
Le
af
Disea
se
Im
age
Ret
r
ie
va
l
Us
ing
Color
Mom
ent
s.
IAE
S
Inte
rnational
Journal
of
Artifi
ci
al
Intelli
g
ence (
IJ
-
AI)
.
2013
;
2
(
1
):
36
-
42
.
[6]
B.
Li
,
e
t
al
.
“
H
y
per
sp
ec
tr
al
id
e
nti
ficat
ion
of
rice
disea
ses
and
pests
base
d
on
princ
ip
al
component
anal
y
s
is
a
nd
proba
bil
ist
ic ne
u
ra
l
n
et
work
,
”
Tr
ansacti
ons of
the CSA
E
,
,
2009
;
5
:
143
–
147.
[7]
H.
Jie,
“
Applicati
on
of
PC
A
Method
on
Pest
Inform
at
ion
Detect
ion
of
Elec
troni
c
Nos
e
,
”
I
EE
E
internation
al
Confe
renc
e
on
I
nformation
A
cqu
isit
ion
.
W
ei
h
ai
,
Shandong,
Chin
a
,
2006
;1456
-
14
68
.
[8]
M
.
M.
Seinl
,
et
al
.
“
Authentic
a
ti
ons
of
M
y
a
nm
ar
Nati
on
al
Reg
istra
ti
on
C
ard
,
”
Indone
sian
Journal
of
Elec
tri
cal
Engi
ne
ering
and
Informatic
s (
IJ
EE
I)
,
2013
;
2
:
53
-
58.
[9]
Prof.H.M.
Deshm
ukh,
Jadha
v
Sanji
v
ani
,
Loha
r
Utkar
sha,
Bhag
a
t
Madhuri,
Salun
ke
Shubhangi.
“
Plant
Leaf
Disea
se
Id
ent
if
ic
a
ti
on
S
ystem
for
Android
,
”
Int
ernati
onal
Journal
of
Ad
va
nce
d
R
ese
arch
i
n
Compute
r
and
Comm
unic
ati
on
Engi
ne
ering.
2
016
;
5
:
396
-
398.
[10]
Sufi
y
an
K
Shekh,
Aniket
B
aitul
e,
Mil
ind
Naret
he,
Sangapp
a
M
al
l
ad,
W
aghda
r
i
kar
,
Dr
D
Y
Pat
il
.
“
Det
ec
t
ion
of
Le
af
Disea
ses
a
nd
Monitori
ng
t
he
Agric
u
lt
ura
l
Resourc
es
using
Android
App
,
”
Inte
rnational
Jo
urnal
of
Inno
vative
Re
search in Co
mputer
and
Com
municat
ion
Enginee
ring
.
2015;
3
:
9540
-
9547.
[11]
Mr.Sus
hant
.
S.C
hava
n,
Mr.Ni
te
s
h.
P.Satr
e
,
Mr.R
aj
a
t.
R.
Deshm
an
e,
Mr.Suhas.B
.
Katka
r.
“
Androi
d
Based
App
to
Preve
nt
Cr
op
Disea
sed
in
Vari
ous
Sea
sons
,
”
Inte
rnat
ional
R
ese
arch
Journal
o
f
Engi
ne
ering
a
nd
Technol
ogy(
IR
J
ET)
.
2016;
3
:
105
2
-
1055.
[12]
Pri
y
ank
a
L
.
Kam
ble
,
Anj
al
i
C
.
Pis
e.
“
Revi
ew
on
Agric
ult
ur
al
Pla
nt
Disea
se
Detect
ion
b
y
using
I
m
age
Proce
ss
ing
.
Inte
rnational
Jo
urnal
of
Lat
est
T
rends i
n
Eng
inee
ring a
nd
Techno
logy
(
IJ
LT
ET)
.
2016;
7
:
335
-
339.
[13]
Stefa
n
L
eutene
g
ger
,
Marg
ari
t
a
C
hli
and
Rola
nd
Y.Sie
gwart
.
“
BRISK:
Bi
nary
Rob
ust I
nvariant
S
c
alabl
e
K
ey
poin
ts
,
”
IEE
E
Internat
ion
al
Conf
ere
n
ce o
n
Com
pute
r
Visi
on,
ICCV 2011
,
Barc
e
lona,
Spai
n,
.
2011;
2548
-
2
555.
Evaluation Warning : The document was created with Spire.PDF for Python.