I
AE
S
I
n
t
e
r
n
at
ion
al
Jou
r
n
al
of
Ar
t
if
icial
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
Vol.
14
,
No.
2
,
Apr
il
20
25
,
pp.
90
7
~
91
6
I
S
S
N:
2252
-
8938
,
DO
I
:
10
.
11591/i
jai
.
v
14
.i
2
.
pp
90
7
-
916
907
Jou
r
n
al
h
omepage
:
ht
tp:
//
ij
ai
.
iaes
c
or
e
.
c
om
Qu
al
ity an
d
s
h
e
lf
-
li
f
e
p
r
e
d
ic
t
io
n
of
c
a
u
li
f
lo
w
e
r
u
si
n
g
m
ac
h
i
n
e
le
ar
n
i
n
g u
n
d
e
r
va
c
u
u
m
an
d
m
od
ifie
d
at
m
os
p
h
e
r
e
p
a
c
k
agi
n
g
M
d
.
Apu
Hos
e
n
,
S
ye
d
M
d
.
Gal
i
b
D
e
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ie
nc
e
a
nd E
ngi
ne
e
r
in
g,
F
a
c
ul
ty
of
E
ngi
ne
e
r
in
g a
nd T
e
c
hnol
ogy
,
J
a
s
hor
e
U
ni
ve
r
s
it
y of
S
c
ie
nc
e
a
nd T
e
c
hnol
ogy, J
a
s
hor
e
, B
a
ngl
a
de
s
h
Ar
t
icle
I
n
f
o
AB
S
T
RA
CT
A
r
ti
c
le
h
is
tor
y
:
R
e
c
e
ived
M
a
r
7,
2024
R
e
vis
e
d
Nov
11,
2024
Ac
c
e
pted
Nov
24,
202
4
E
n
s
u
r
i
n
g
t
h
e
fres
h
n
e
s
s
a
n
d
q
u
al
i
t
y
o
f
ca
u
l
i
fl
o
w
er
d
u
r
i
n
g
s
t
o
ra
g
e
an
d
t
ran
s
p
o
rt
a
t
i
o
n
i
s
es
s
en
t
i
a
l
d
u
e
t
o
i
t
s
h
i
g
h
p
eri
s
h
a
b
i
l
i
t
y
.
T
h
i
s
s
t
u
d
y
h
arn
e
s
s
e
s
t
h
e
p
o
w
er
o
f
mach
i
n
e
l
ear
n
i
n
g
t
o
p
red
i
ct
t
h
e
q
u
al
i
t
y
an
d
s
h
e
l
f
-
l
i
fe
o
f
cau
l
i
fl
o
w
er
u
n
d
er
co
s
t
-
eff
ect
i
v
e
v
ac
u
u
m
an
d
m
o
d
i
fi
e
d
at
mo
s
p
h
ere
p
ack
a
g
i
n
g
(M
A
P)
t
ec
h
n
i
q
u
es
.
By
i
n
v
es
t
i
g
at
i
n
g
k
e
y
p
ar
amet
ers
s
u
c
h
as
t
o
t
al
s
o
l
u
b
l
e
s
o
l
i
d
s
(
T
SS),
p
H
,
w
ei
g
h
t
l
o
s
s
,
an
d
co
l
o
r
ch
a
n
g
e,
a
s
i
g
n
i
fi
ca
n
t
i
m
p
act
o
n
p
o
s
t
-
p
ack
a
g
i
n
g
q
u
a
l
i
t
y
w
a
s
i
d
en
t
i
f
i
ed
.
T
o
ad
d
re
s
s
t
h
e
c
h
al
l
en
g
e
o
f
accu
ra
t
e
c
o
l
o
r
ch
a
n
g
e
me
as
u
remen
t
,
an
i
n
n
o
v
a
t
i
v
e
met
h
o
d
u
t
i
l
i
zi
n
g
a
b
i
l
at
era
l
fi
l
t
er
fo
r
n
o
i
s
e
red
u
ct
i
o
n
an
d
p
art
i
cl
e
s
w
arm
o
p
t
i
m
i
zat
i
o
n
(PSO
)
w
i
t
h
Mark
o
v
ran
d
o
m
fi
el
d
(MRF)
s
eg
men
t
a
t
i
o
n
w
a
s
d
e
v
el
o
p
e
d
.
T
SS,
w
ei
g
h
t
l
o
s
s
,
an
d
c
o
l
o
r
ch
an
g
e
w
ere
i
d
en
t
i
f
i
ed
a
s
k
ey
p
a
rame
t
ers
,
a
n
d
l
ev
era
g
i
n
g
t
h
e
s
e
p
arame
t
ers
,
ar
t
i
f
i
ci
a
l
n
eu
ra
l
n
e
t
w
o
rk
s
(
A
N
N
)
w
er
e
emp
l
o
y
e
d
t
o
crea
t
e
h
i
g
h
l
y
p
reci
s
e
p
red
i
ct
i
v
e
m
o
d
e
l
s
,
ac
h
i
e
v
i
n
g
R
-
s
q
u
are
d
v
al
u
es
o
f
0
.
9
5
2
f
o
r
T
SS,
0
.
9
9
2
f
o
r
w
e
i
g
h
t
l
o
s
s
,
a
n
d
0
.
9
8
1
fo
r
co
l
o
r
c
h
an
g
e.
T
h
i
s
ap
p
ro
ac
h
n
o
t
o
n
l
y
e
n
h
a
n
ces
t
h
e
eff
i
ci
e
n
cy
a
n
d
s
u
s
t
a
i
n
a
b
i
l
i
t
y
o
f
f
o
o
d
p
ro
d
u
ct
i
o
n
an
d
d
i
s
t
ri
b
u
t
i
o
n
b
u
t
a
l
s
o
mi
n
i
m
i
zes
f
o
o
d
w
a
s
t
e
a
n
d
max
i
m
i
zes
p
ro
f
i
t
a
b
i
l
i
t
y
f
o
r
cau
l
i
fl
o
w
er
i
n
g
l
o
b
al
mar
k
et
s
t
h
r
o
u
g
h
t
h
e
u
s
e
o
f
c
o
s
t
-
effect
i
v
e
p
ac
k
ag
i
n
g
s
o
l
u
t
i
o
n
s
.
K
e
y
w
o
r
d
s
:
M
a
c
hine
lea
r
ning
M
od
i
f
ie
d
a
tm
os
p
he
r
e
pa
c
ka
gi
ng
P
a
r
ti
c
le
s
wa
r
m
opti
mi
z
a
ti
on
Qua
li
ty
a
s
s
e
s
s
ment
S
he
lf
-
li
f
e
pr
e
diction
Va
c
uum
pa
c
ka
ging
Th
i
s
i
s
a
n
o
p
en
a
c
ces
s
a
r
t
i
c
l
e
u
n
d
e
r
t
h
e
CC
B
Y
-
SA
l
i
ce
n
s
e.
C
or
r
e
s
pon
din
g
A
u
th
or
:
Md
.
Apu
Hos
e
n
De
pa
r
tm
e
nt
of
C
omput
e
r
S
c
ienc
e
a
nd
E
nginee
r
ing
,
F
a
c
ult
y
of
E
nginee
r
ing
a
nd
T
e
c
hnology
J
a
s
hor
e
Unive
r
s
it
y
of
S
c
ienc
e
a
nd
T
e
c
hnology
J
a
s
hor
e
-
7408,
B
a
nglade
s
h
E
mail:
a
pu
.
c
s
e
.
jus
t@gm
a
il
.
c
om
1.
I
NT
RODU
C
T
I
ON
C
a
uli
f
lowe
r
r
a
nks
a
mong
the
mos
t
s
igni
f
ica
nt
c
r
o
ps
wor
ldwide
in
ter
ms
of
nutr
it
ion,
va
lued
f
or
it
s
ve
r
s
a
ti
li
ty
in
c
uli
na
r
y
a
ppli
c
a
ti
ons
,
including
s
a
lads
a
nd
va
r
ious
c
ooke
d
dis
h
e
s
.
I
ts
he
a
lt
h
be
ne
f
it
s
,
s
uc
h
a
s
c
a
nc
e
r
r
e
s
is
tanc
e
a
nd
de
toxi
f
ica
ti
on
pr
ope
r
ti
e
s
,
unde
r
s
c
or
e
it
s
im
por
tanc
e
in
moder
n
diets
[
1]
,
[
2
]
.
I
t
is
a
ls
o
a
good
s
our
c
e
of
vit
a
mi
ns
B
,
C
,
E
,
a
nd
K,
dieta
r
y
f
iber
,
f
o
li
c
a
c
id,
omega
-
3
f
a
tt
y
a
c
ids
,
pr
oteins
,
pho
s
phor
us
,
potas
s
ium
,
ir
on
,
magne
s
ium
,
a
nd
manga
ne
s
e
[
3]
,
[
4]
.
C
ons
e
que
ntl
y,
c
a
uli
f
lowe
r
ha
s
gr
e
a
t
de
mand
globally.
A
s
igni
f
ica
nt
qua
nti
ty
of
c
a
uli
f
lowe
r
is
c
ult
ivate
d
in
B
a
nglade
s
h,
de
mons
tr
a
ti
ng
it
s
c
a
pa
c
it
y
t
o
f
ulf
il
l
domes
ti
c
ne
e
ds
a
nd
pa
r
ti
c
ipate
in
pr
of
it
a
ble
e
xpor
t
mar
ke
ts
[
5]
.
How
e
ve
r
,
c
a
uli
f
lowe
r
e
xhib
it
s
high
pe
r
is
ha
bil
it
y
pos
t
-
ha
r
ve
s
t
due
to
it
s
e
l
e
va
ted
r
e
s
pir
a
ti
on
r
a
te
a
nd
pr
ope
ns
it
y
f
or
wa
ter
los
s
[
6]
–
[
8]
.
T
o
mi
ti
ga
te
thes
e
de
f
e
c
ts
without
c
ompr
omi
s
ing
nutr
it
ional
qua
li
ty
,
va
r
ious
methods
ha
ve
be
e
n
e
mpl
oye
d,
including
low
-
tempe
r
a
t
ur
e
s
tor
a
ge
[
9
]
,
pa
c
ka
ging
[
10
]
,
s
a
nit
iza
ti
on
[
11]
,
a
n
ti
-
br
owning
dipp
ing
tr
e
a
tm
e
nts
[
12]
,
a
nd
a
ppl
ica
ti
on
of
e
dibl
e
c
oa
ti
ngs
[
13
]
.
F
or
s
uc
c
e
s
s
f
ul
e
xpor
t,
it
is
c
r
uc
ial
to
de
ter
mi
ne
how
long
the
qua
li
ty
of
the
p
r
oduc
ts
will
be
p
r
e
s
e
r
ve
d
unde
r
s
p
e
c
if
ic
methods
,
mea
ning
the
s
he
lf
li
f
e
a
nd
qua
li
ty
mus
t
be
a
c
c
ur
a
tely
a
s
s
e
s
s
e
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
2,
Apr
il
2025
:
907
-
9
16
908
Ac
c
ur
a
te
s
he
lf
-
li
f
e
pr
e
diction
he
lps
e
xpor
ter
s
mee
t
the
qua
li
ty
s
tanda
r
ds
a
nd
r
e
gulations
of
im
por
ti
ng
c
ountr
ies
,
the
r
e
by
a
voidi
ng
r
e
jec
ti
ons
a
nd
f
inanc
ial
los
s
e
s
.
I
t
e
ns
ur
e
s
t
ha
t
the
c
a
uli
f
lowe
r
maintains
it
s
nutr
it
ional
va
lue,
a
ppe
a
r
a
nc
e
,
a
nd
tas
te
dur
i
ng
tr
a
ns
it
.
Additi
ona
ll
y
,
it
a
ll
ows
e
xpor
ter
s
to
opti
mi
z
e
pa
c
ka
ging,
s
tor
a
ge
,
a
nd
tr
a
ns
por
tation
methods
,
r
e
duc
ing
wa
s
te
a
nd
incr
e
a
s
ing
pr
of
it
a
bil
i
ty.
T
his
a
ls
o
he
lps
buil
d
tr
us
t
with
int
e
r
na
ti
ona
l
buye
r
s
,
who
e
xpe
c
t
r
e
li
a
ble
a
nd
c
ons
is
tent
qua
li
ty
in
the
pr
oduc
e
they
r
e
c
e
ive.
Ulti
mate
ly,
a
c
c
ur
a
te
s
he
lf
-
li
f
e
pr
e
diction
lea
ds
to
incr
e
a
s
e
d
r
e
ve
nue
f
or
the
c
ountr
y
a
nd
a
s
tr
onge
r
c
ompetit
ive
pos
it
ion
in
the
global
ma
r
ke
t.
T
o
main
tain
qua
li
ty
a
nd
e
xtend
s
he
lf
li
f
e
dur
ing
e
xpor
t
,
pa
c
ka
ging
is
the
mos
t
popular
method
.
Among
the
va
r
ious
pa
c
ka
ging
tec
hniques
,
modi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
ging
(
M
AP)
a
nd
va
c
uum
pa
c
k
a
ging
a
r
e
the
mos
t
c
omm
only
us
e
d.
Va
c
uum
pa
c
ka
ging
inv
olves
r
e
movi
ng
a
ir
f
r
om
the
pa
c
ka
ge
be
f
o
r
e
s
e
a
li
ng
[
14]
,
while
M
AP
manipulate
s
the
a
tm
os
phe
r
e
in
s
ide
the
pa
c
ka
ging
[
15]
.
I
n
M
AP
,
d
if
f
e
r
e
nt
ga
s
r
a
ti
os
li
ke
nit
r
oge
n
(N
2
)
,
oxyge
n
(
O
2
)
,
a
nd
c
a
r
bon
dioxi
de
(
C
O
2
)
a
r
e
c
omm
only
us
e
d
to
modi
f
y
the
a
t
mos
phe
r
e
[
16]
–
[
18]
.
How
e
ve
r
,
ga
s
r
a
ti
o
-
ba
s
e
d
M
AP
is
c
os
tl
y
[
19]
.
R
e
s
e
a
r
c
he
r
s
ha
ve
c
onduc
ted
s
tudi
e
s
to
pr
e
dict
qua
li
ty
a
nd
s
he
lf
li
f
e
unde
r
ga
s
r
a
ti
o
-
ba
s
e
d
M
AP.
Due
to
the
high
c
os
t
of
ga
s
r
a
ti
o
-
ba
s
e
d
pa
c
ka
ging,
thi
s
r
e
s
e
a
r
c
h
f
oc
us
e
d
on
c
he
mi
c
a
l
-
ba
s
e
d
M
AP
pa
c
ka
ging
a
nd
va
c
uum
pa
c
ka
ging
tec
hniques
to
r
e
duc
e
c
os
ts
a
nd
identif
y
r
e
leva
nt
f
a
c
tor
s
to
pr
e
dict
qua
li
ty
a
nd
s
he
lf
li
f
e
.
Unde
r
s
tanding
the
f
a
c
tor
s
that
inf
luenc
e
the
de
gr
a
da
ti
on
of
c
a
uli
f
lowe
r
qua
li
ty
is
e
s
s
e
nti
a
l
f
or
s
uc
h
pr
e
dictio
ns
.
P
a
r
a
mete
r
s
s
uc
h
a
s
c
olo
r
c
ha
nge
,
we
ight
los
s
,
pH,
a
nd
tot
a
l
s
olubl
e
s
oli
ds
(
T
S
S
)
a
r
e
e
va
luate
d
to
a
s
s
e
s
s
q
ua
li
ty
a
lt
e
r
a
ti
ons
.
Among
thes
e
,
c
olo
r
c
ha
nge
mea
s
ur
e
ment
is
a
c
r
it
ica
l
f
a
c
tor
.
T
r
a
dit
ional
s
e
gmenta
ti
on
me
thods
ha
ve
be
e
n
us
e
d
to
mea
s
ur
e
c
olor
c
ha
nge
,
but
in
c
a
uli
f
lowe
r
,
s
mall
blank
s
pa
c
e
s
withi
n
the
f
lo
r
e
t
s
c
a
n
lea
d
to
inac
c
ur
a
te
mea
s
ur
e
ments
whe
n
im
a
ge
s
a
r
e
c
a
ptur
e
d.
T
o
a
ddr
e
s
s
thi
s
pr
oblem,
thi
s
s
tudy
pr
o
pos
e
d
a
n
opti
mi
z
e
d
c
olo
r
c
ha
nge
mea
s
ur
e
ment
s
ys
tem
to
im
pr
ove
a
c
c
ur
a
c
y.
T
he
ke
y
c
ontr
ibut
ion
s
of
thi
s
r
e
s
e
a
r
c
h
include
:
i)
inves
ti
ga
te
c
os
t
-
e
f
f
e
c
ti
ve
pa
c
ka
ging
tec
hniques
f
or
c
a
uli
f
lowe
r
a
nd
a
c
quir
e
the
pa
c
k
a
ge
d
da
ta
;
ii
)
identi
f
ica
ti
on
o
f
s
igni
f
ica
nt
pa
r
a
m
e
ter
s
f
or
pr
e
dicting
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
he
lf
-
li
f
e
pos
t
-
pa
c
ka
ging,
ba
s
e
d
on
c
omp
r
e
he
ns
ive
tes
ti
ng
a
f
ter
pa
c
ka
ging
;
ii
i)
de
ve
lopm
e
nt
of
a
nove
l
s
ys
tem
f
or
e
f
f
icie
ntl
y
mea
s
ur
ing
c
olor
c
ha
nge
in
c
a
uli
f
lowe
r
pos
t
-
pa
c
ka
ging
;
a
nd
iv)
a
c
c
ur
a
tely
f
o
r
e
c
a
s
t
the
qua
li
ty
a
nd
s
he
lf
li
f
e
of
c
a
uli
f
lowe
r
.
T
he
r
e
mainde
r
o
f
the
pa
pe
r
is
o
r
ga
nize
d
int
o
s
e
ve
r
a
l
s
e
c
ti
ons
.
S
e
c
ti
on
2
pr
ovides
a
r
e
view
of
the
r
e
leva
nt
li
ter
a
tur
e
,
s
e
tt
ing
the
f
ounda
ti
on
f
or
th
e
s
tudy.
S
e
c
ti
on
3
de
tails
the
methodology
us
e
d
in
the
r
e
s
e
a
r
c
h,
while
s
e
c
ti
on
4
pr
e
s
e
nts
the
r
e
s
ult
s
a
nd
e
nga
ge
s
in
a
dis
c
us
s
ion
of
the
f
indi
ngs
.
F
inally,
s
ec
ti
on
5
c
onc
ludes
the
pa
pe
r
by
s
umm
a
r
izing
ke
y
ins
ight
s
a
nd
s
ugge
s
ti
ng
dir
e
c
ti
ons
f
or
f
utur
e
r
e
s
e
a
r
c
h.
2.
L
I
T
E
RA
T
UR
E
RE
VI
E
W
R
e
s
e
a
r
c
h
f
oc
us
e
d
on
pr
e
dicting
the
s
he
lf
li
f
e
of
pa
c
ka
ge
d
c
a
uli
f
lowe
r
is
r
e
latively
unc
omm
on
in
the
s
c
ientif
ic
c
omm
unit
y.
How
e
ve
r
,
r
e
s
e
a
r
c
he
r
s
ha
ve
made
s
igni
f
ica
nt
s
tr
ides
in
f
o
r
e
c
a
s
ti
ng
the
s
he
lf
li
f
e
of
va
r
ious
pe
r
is
ha
ble
goods
,
s
uc
h
a
s
f
r
uit
s
,
ve
ge
table
s
,
a
nd
f
is
h.
S
o
me
notable
s
tudi
e
s
in
thi
s
a
r
e
a
includ
e
.
M
oha
mm
e
d
e
t
al
.
[
20]
e
mphas
ize
the
s
igni
f
ica
nc
e
of
maintaining
the
s
a
f
e
ty
a
nd
qua
li
ty
of
f
r
e
s
h
f
r
uit
s
by
uti
li
z
ing
a
dva
nc
e
d
tec
hnologi
e
s
li
ke
M
AP.
T
he
ir
s
tudy
int
r
oduc
e
s
a
c
os
t
-
e
f
f
e
c
ti
ve
method
e
mpl
oying
ti
ny
mac
hine
lea
r
ning
(
T
inyM
L
)
a
nd
mul
ti
s
pe
c
tr
a
l
s
e
ns
or
s
to
pr
e
dict
the
qu
a
li
ty
pa
r
a
mete
r
s
a
nd
s
he
lf
li
f
e
of
pa
c
ka
ge
d
f
r
e
s
h
da
tes
unde
r
dif
f
e
r
e
nt
c
ondit
ions
.
T
he
f
indi
ngs
de
mons
tr
a
te
a
s
ubs
tantial
incr
e
a
s
e
in
s
he
lf
li
f
e
,
pa
r
ti
c
ular
ly
with
va
c
uum
a
nd
M
AP1
pa
c
ka
ging,
with
high
pr
e
diction
a
c
c
ur
a
c
y
(
R
s
qua
r
e
d
va
lue,
R
2
=
0.
951)
.
T
h
r
ough
a
n
op
ti
mal
ne
ur
a
l
ne
twor
k
model,
va
r
ious
qua
li
ty
pa
r
a
mete
r
s
s
uc
h
a
s
pH,
T
S
S
,
s
uga
r
c
ontent,
mo
is
tur
e
c
ontent
(
M
C
)
,
a
nd
tannin
c
ont
e
nt
we
r
e
e
f
f
icie
ntl
y
p
r
e
dicte
d.
T
he
s
e
models
of
f
e
r
r
obus
t
tool
s
f
or
a
s
s
e
s
s
ing
f
r
uit
qua
l
it
y
a
c
c
ur
a
tely,
ther
e
by
be
ne
f
it
ing
p
r
oduc
e
r
s
a
nd
c
ons
umer
s
in
opti
mi
z
in
g
s
upply
c
ha
in
mana
ge
ment
a
nd
e
ns
ur
ing
f
r
e
s
h
f
r
uit
qua
li
ty.
Albe
r
t
-
W
e
is
s
a
nd
Os
man
[
21
]
f
oc
us
on
a
s
s
e
s
s
in
g
a
gr
icultur
a
l
pr
oduc
t
qua
li
ty
a
nd
r
ipene
s
s
us
ing
non
-
de
s
tr
uc
ti
ve
tes
ti
ng
tec
hnique
s
,
s
pe
c
i
f
ica
ll
y
a
c
ous
ti
c
tes
ti
ng.
T
he
y
a
ddr
e
s
s
c
ha
ll
e
nge
s
a
s
s
o
c
ia
ted
with
e
mpl
oying
de
e
p
lea
r
ning
(
DL
)
methods
li
ke
c
onvolut
ional
ne
ur
a
l
ne
twor
ks
(
C
NN
s
)
due
to
da
ta
inef
f
icie
nc
y
a
nd
a
lac
k
of
a
nnotate
d
da
ta.
T
o
tac
kle
thes
e
c
ha
ll
e
nge
s
,
the
s
tudy
int
r
oduc
e
s
a
c
ti
ve
lea
r
ning
a
s
a
f
r
a
mew
or
k,
pa
r
ti
c
ular
ly
r
e
leva
nt
whe
n
labe
led
ins
tanc
e
s
a
r
e
s
c
a
r
c
e
.
T
he
y
pr
opos
e
the
k
-
de
ter
mi
na
ntal
point
pr
oc
e
s
s
e
s
(k
-
DPP
)
method
withi
n
the
a
c
ti
ve
lea
r
ning
f
r
a
mew
or
k,
whic
h
a
im
s
to
e
nha
nc
e
e
xplor
a
ti
on
withi
n
t
he
f
e
a
tur
e
s
pa
c
e
by
s
e
lec
ti
ng
a
diver
s
e
s
ubs
e
t
k
.
T
his
a
ppr
o
a
c
h
de
mons
tr
a
tes
e
f
f
icie
nc
y,
e
s
pe
c
ially
in
s
c
e
na
r
i
os
with
li
mi
ted
labe
led
s
a
mpl
e
s
,
a
c
hieving
a
n
a
c
c
ur
a
c
y
o
f
73.
91%
in
gr
a
ding
'Ga
li
a
'
mus
kmelons
ba
s
e
d
on
s
he
lf
li
f
e
.
I
n
a
s
tudy
led
by
I
or
l
iam
e
t
al
.
[
22]
,
mac
hine
lea
r
n
ing
tec
hniques
,
includ
ing
s
uppor
t
ve
c
tor
mac
hine
,
n
a
ïv
e
B
a
ye
s
,
de
c
is
i
on
tr
e
e
,
logi
s
ti
c
r
e
gr
e
s
s
ion,
a
n
d
k
-
ne
a
r
e
s
t
ne
ighbor
a
lgor
it
hms
,
a
r
e
a
ppli
e
d
to
p
r
e
dict
the
s
he
lf
li
f
e
of
Ok
r
a
.
T
he
r
e
s
e
a
r
c
h
a
im
s
to
mi
ti
ga
te
potential
ha
r
m
a
s
s
oc
iate
d
with
c
ons
umi
ng
Okr
a
be
yond
it
s
s
he
lf
li
f
e
.
Va
r
ious
pa
r
a
mete
r
s
s
uc
h
a
s
we
ight
lo
s
s
,
f
ir
mn
e
s
s
,
ti
tr
a
ble
a
c
id,
T
S
S
,
vit
a
mi
n
C
/As
c
or
bic
a
c
id
c
ontent,
a
nd
pH
a
r
e
uti
li
z
e
d
a
s
input
s
f
or
thes
e
m
a
c
hine
-
lea
r
ning
models
.
Nota
bly,
s
uppor
t
ve
c
tor
mac
hine
,
n
a
ïve
B
a
ye
s
,
a
nd
de
c
is
ion
tr
e
e
a
lgor
it
hms
a
c
hieve
pe
r
f
e
c
t
pr
e
dictions
of
Okr
a
's
s
he
lf
li
f
e
,
e
a
c
h
wit
h
a
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Qualit
y
and
s
he
lf
-
li
fe
pr
e
diction
of
c
auli
fl
ow
e
r
us
i
ng
mac
hine
lear
ning
unde
r
v
ac
uum
…
(
M
d
A
pu
H
os
e
n
)
909
a
c
c
ur
a
c
y
of
100
%
.
L
ogis
ti
c
r
e
gr
e
s
s
ion
a
nd
k
-
ne
a
r
e
s
t
ne
ighbor
a
lgor
it
hms
a
c
hieve
s
li
ghtl
y
lowe
r
a
c
c
u
r
a
c
ies
a
t
88.
89
a
nd
88.
33%
,
r
e
s
pe
c
ti
ve
ly.
T
he
s
tudy
c
onc
ludes
that
mac
hine
lea
r
ning
tec
hniques
,
pa
r
ti
c
ular
ly
s
uppor
t
ve
c
tor
mac
hine
,
n
a
ïve
B
a
ye
s
,
a
nd
de
c
is
ion
tr
e
e
,
a
r
e
e
f
f
e
c
ti
ve
in
a
c
c
ur
a
tely
pr
e
dicting
the
s
he
lf
li
f
e
of
Okr
a
.
Alde
n
e
t
al
.
[
23
]
inves
ti
ga
te
the
im
pa
c
t
o
f
M
AP
on
c
a
uli
f
lowe
r
s
he
lf
li
f
e
a
nd
qua
li
ty
.
T
he
ir
s
tudy
a
na
lyze
s
f
our
pa
c
ka
ging
methods
ove
r
30
da
ys
,
de
mons
tr
a
ti
ng
that
M
AP1
s
igni
f
ica
ntl
y
e
xtends
s
he
lf
li
f
e
be
yond
30
da
ys
.
Us
ing
a
r
ti
f
icia
l
ne
ur
a
l
ne
twor
ks
(
AN
N)
,
a
model
with
one
hidden
laye
r
a
nd
12
ne
ur
ons
a
c
c
ur
a
tely
pr
e
dicts
c
a
uli
f
lowe
r
s
he
lf
li
f
e
ba
s
e
d
on
c
olor
c
ha
nge
s
,
e
xhibi
ti
ng
high
a
c
c
ur
a
c
y
with
a
mea
n
s
qua
r
e
e
r
r
or
of
0.
0095
a
nd
R
s
qua
r
e
d
va
lue
(
R
2
)
of
0.
990
.
C
a
uli
f
lowe
r
pa
c
ke
d
with
M
AP1
de
mo
ns
tr
a
tes
mar
ke
ti
ng
c
a
pa
bil
it
y
f
or
up
to
50
da
ys
in
ter
ms
of
tot
a
l
c
olor
c
ha
nge
s
.
M
e
c
ha
nica
l
pr
ope
r
ti
e
s
s
h
owe
d
no
s
igni
f
ica
nt
dif
f
e
r
e
nc
e
s
a
mong
pa
c
ka
ging
methods
on
da
ys
20,
25,
a
nd
30,
whi
le
c
olor
c
ha
nge
s
a
nd
we
ight
los
s
e
xhibi
ted
s
igni
f
ica
nt
dif
f
e
r
e
nc
e
s
in
thes
e
c
ompar
is
ons
.
F
u
e
t
al
.
[
24
]
inves
ti
ga
te
the
s
he
lf
li
f
e
of
tr
icholo
m
a
mats
utake
(
T
.
mats
utake
)
f
r
om
T
ibet,
f
oc
us
ing
on
M
AP
c
ondit
ions
in
a
c
old
c
ha
in
to
unde
r
s
tand
qua
li
ty
c
ha
n
ge
s
dur
ing
T
.
mats
utake
's
s
he
lf
li
f
e
.
T
h
e
ir
s
tudy
a
na
lyze
s
ke
y
qua
li
ty
indi
c
a
tor
s
s
uc
h
a
s
ha
r
dne
s
s
,
c
olor
,
odor
,
pH
,
s
olubl
e
s
oli
ds
c
ontent
(
S
S
C
)
,
a
n
d
M
C
a
t
s
pe
c
if
ic
e
nvir
onmenta
l
c
ondit
ions
.
T
he
s
e
ns
or
y
e
v
a
luation
highl
ight
s
odor
s
e
ns
it
ivi
ty
a
s
a
f
r
e
s
hne
s
s
i
ndica
tor
.
P
hys
iol
ogica
l
c
ha
nge
s
in
pH,
S
S
C
,
a
nd
M
C
a
r
e
c
a
tegor
ize
d
int
o
th
r
e
e
pe
r
iods
,
r
e
f
lec
ti
ng
c
a
p
s
pr
e
a
d,
gr
a
dua
l
c
ha
nge
s
,
a
nd
c
ompl
ica
ted
de
ter
io
r
a
ti
on.
T
he
s
tudy
e
s
tablis
he
s
a
ba
c
k
pr
opa
ga
ti
on
(
B
P
)
ne
ur
a
l
ne
twor
k
model
to
pr
e
dict
r
e
maining
s
he
lf
l
if
e
ba
s
e
d
on
qua
li
ty
indi
c
a
tor
s
,
opti
mi
z
ing
thr
ough
c
or
r
e
lation
a
na
ly
s
is
.
T
his
r
e
s
e
a
r
c
h
is
a
nti
c
ipate
d
to
be
ne
f
it
the
tr
a
ns
por
tatio
n
a
nd
pr
e
s
e
r
va
ti
on
of
T
.
mats
utake
,
r
e
duc
ing
los
s
e
s
in
the
pos
thar
ve
s
t
c
ha
in.
T
he
r
e
view
e
d
li
ter
a
tur
e
unde
r
s
c
or
e
s
the
s
igni
f
ica
nc
e
of
pr
e
dicting
s
he
lf
li
f
e
in
the
domain
of
pe
r
is
ha
ble
goods
,
pa
r
ti
c
ula
r
ly
f
r
uit
s
,
ve
ge
table
s
,
a
nd
f
is
h.
C
oll
e
c
ti
ve
ly,
thes
e
s
tudi
e
s
c
ontr
ibut
e
to
the
a
dva
nc
e
ment
of
pr
e
dictive
models
a
nd
p
r
e
s
e
r
va
ti
on
tec
hniques
.
T
his
ul
ti
mate
ly
e
nha
nc
ing
f
ood
s
a
f
e
ty,
qua
li
ty,
a
nd
s
upply
c
ha
in
mana
ge
ment
in
the
pe
r
is
ha
ble
goods
indus
tr
y.
3.
M
E
T
HO
DOL
OG
Y
T
o
a
c
c
ur
a
tely
pr
e
dict
the
qua
li
ty
a
nd
s
he
lf
-
li
f
e
o
f
c
a
uli
f
lowe
r
unde
r
va
c
uum
a
nd
M
AP,
a
s
ys
tema
ti
c
a
ppr
oa
c
h
invol
ving
mul
ti
ple
s
tage
s
wa
s
e
mpl
oy
e
d.
T
his
c
ompr
e
he
ns
ive
pr
oc
e
s
s
is
de
picte
d
in
the
f
low
diagr
a
m
of
F
igur
e
1,
whic
h
ou
tl
ines
e
a
c
h
c
r
it
ica
l
s
tep
in
the
methodology.
E
a
c
h
s
tage
is
e
s
s
e
nti
a
l
to
e
ns
ur
e
the
pr
e
c
is
ion
a
nd
r
e
li
a
bil
i
ty
of
the
p
r
e
dictive
mode
ls
.
F
igur
e
1.
W
or
kf
low
diagr
a
m
of
the
s
ys
tem
3.
1.
P
ac
k
age
d
d
at
a
ac
q
u
is
it
io
n
T
h
e
e
xp
e
r
i
men
ta
l
a
pp
r
oa
c
h
wa
s
me
ti
c
u
lo
us
l
y
de
s
ig
ne
d
to
in
ve
s
ti
ga
te
th
e
in
f
lue
nc
e
o
f
di
f
f
e
r
e
n
t
p
a
c
ka
g
i
ng
m
e
t
ho
ds
o
n
th
e
q
ua
l
i
ty
o
f
c
a
u
li
f
lo
we
r
.
S
pe
c
i
f
ica
ll
y
,
t
he
f
o
c
us
w
a
s
on
tw
o
p
r
i
ma
r
y
te
c
h
n
iq
ue
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
2,
Apr
il
2025
:
907
-
9
16
910
v
a
c
uu
m
p
a
c
ka
g
in
g
a
n
d
M
A
P
.
As
t
he
p
r
i
ma
r
y
o
bj
e
c
ti
ve
o
f
M
AP
is
t
o
a
lt
e
r
t
he
a
t
mos
ph
e
r
e
wi
th
in
t
he
pa
c
ka
gi
ng
,
a
c
he
m
ica
l
-
ba
s
e
d
a
pp
r
oa
c
h
w
a
s
a
p
pl
ie
d
.
KM
nO
₄
,
C
a
O
,
a
n
d
a
c
t
iva
te
d
c
a
r
bo
n
we
r
e
us
e
d
be
c
a
us
e
KM
nO₄
a
bs
o
r
bs
e
th
y
len
e
t
o
s
l
ow
r
ip
e
n
in
g
,
C
a
O
r
e
d
uc
e
s
m
ois
tu
r
e
t
o
p
r
e
v
e
n
t
m
ic
r
ob
ia
l
g
r
ow
th
,
a
nd
a
c
ti
va
te
d
c
a
r
bo
n
a
ds
o
r
b
s
u
nw
a
n
te
d
ga
s
e
s
a
nd
od
or
s
to
m
a
i
n
tai
n
f
r
e
s
h
ne
s
s
.
T
o
e
va
lu
a
t
e
th
e
c
o
mb
in
a
t
io
n
o
f
c
h
e
m
ica
ls
,
t
h
r
e
e
d
i
f
f
e
r
e
nt
M
A
P
s
wi
th
va
r
yi
ng
c
h
e
m
ica
l
c
o
mb
in
a
t
io
ns
w
e
r
e
t
e
s
te
d
.
M
A
P
1
u
ti
li
z
e
d
C
a
O
a
nd
a
c
t
i
va
t
e
d
c
a
r
b
on
,
M
A
P
2
e
mp
l
oye
d
KM
nO
₄
a
n
d
a
c
t
iv
a
t
e
d
c
a
r
b
on
,
a
nd
M
AP
3
u
t
il
iz
e
d
KM
nO
₄
,
C
a
O
,
a
nd
a
c
t
iv
a
t
e
d
c
a
r
b
on
.
F
or
the
e
xpe
r
im
e
nt
,
f
r
e
s
h
c
a
uli
f
lowe
r
s
pe
c
im
e
ns
we
r
e
s
our
c
e
d
f
r
om
c
r
op
f
ields
in
J
a
s
hor
e
,
B
a
nglade
s
h,
with
e
a
c
h
s
pe
c
im
e
n
ha
ving
a
n
a
ve
r
a
ge
we
ight
of
650
gr
a
ms
.
T
o
e
ns
ur
e
the
int
e
gr
it
y
of
the
c
r
op
dur
ing
c
oll
e
c
ti
on
a
nd
tr
a
ns
por
tation,
c
a
uli
f
lowe
r
s
we
r
e
ha
r
ve
s
ted
with
their
lea
ve
s
e
nve
lopi
ng
th
e
f
lowe
r
he
a
ds
.
T
he
s
pe
c
im
e
ns
we
r
e
then
r
a
ndoml
y
divi
de
d
int
o
f
ou
r
gr
oups
,
with
e
a
c
h
g
r
oup
a
s
s
igned
to
th
r
e
e
M
AP
a
nd
v
a
c
uum
pa
c
ka
ging
type
to
e
ns
ur
e
unbia
s
e
d
dis
tr
ibut
ion
a
c
r
os
s
e
xpe
r
im
e
ntal
c
ondit
ions
.
B
a
s
e
li
ne
da
ta,
including
ini
ti
a
l
pH,
T
S
S
,
we
ight
,
a
nd
vis
ua
l
c
ha
r
a
c
ter
is
ti
c
s
,
we
r
e
meticulous
ly
r
e
c
or
de
d
us
ing
a
n
im
a
ging
s
ys
tem.
T
he
i
maging
s
ys
tem
wa
s
s
pe
c
if
ica
ll
y
de
s
i
gne
d
a
nd
c
ons
tr
uc
ted
us
ing
a
woode
n
c
ube
.
A
plat
f
or
m
f
or
plac
ing
the
s
a
mpl
e
s
wa
s
pos
it
ioned
a
t
the
c
e
nter
of
the
bott
om
ba
s
e
,
a
nd
to
mi
nim
ize
li
ght
r
e
f
lec
ti
on,
the
int
e
r
ior
s
ur
f
a
c
e
s
of
the
box
we
r
e
pa
int
e
d
b
lac
k.
T
he
c
a
uli
f
lowe
r
wa
s
then
pa
c
ka
ge
d
in
polyethyle
ne
pouc
he
s
with
s
a
c
he
t
s
of
c
he
mi
c
a
ls
.
Af
ter
that,
a
ll
M
AP
-
pa
c
ka
ge
d
c
a
uli
f
lowe
r
s
a
mpl
e
s
we
r
e
s
tor
e
d
in
a
r
e
f
r
iger
a
ted
e
nvir
onment
a
t
4±
1
°C
,
while
va
c
uum
-
pa
c
ka
ge
d
c
a
uli
f
lowe
r
wa
s
s
tor
e
d
a
t
r
oom
tempe
r
a
tur
e
.
Ove
r
a
pe
r
iod
of
s
e
ve
n
da
ys
,
va
r
ious
tes
ts
,
i
nc
ludi
ng
pH
tes
ti
ng,
T
S
S
mea
s
ur
e
ment,
a
nd
we
ight
los
s
a
s
s
e
s
s
ment,
we
r
e
c
onduc
ted
in
a
c
he
mi
c
a
l
la
bor
a
tor
y.
Additi
ona
ll
y,
im
a
ge
s
of
e
a
c
h
s
a
mpl
e
we
r
e
c
a
p
tur
e
d
thr
oughout
the
pa
c
ka
ging
pe
r
iod
.
T
he
s
c
he
matic
r
e
pr
e
s
e
ntation
of
the
da
ta
a
c
quis
it
ion
pr
oc
e
s
s
is
il
l
us
tr
a
ted
in
F
igur
e
2.
F
igur
e
2.
Dia
gr
a
m
of
the
pa
c
ka
ge
d
da
ta
a
c
quis
it
ion
pr
oc
e
dur
e
3.
2
.
Color
m
e
as
u
r
e
m
e
n
t
s
ys
t
e
m
d
e
ve
lop
m
e
n
t
M
e
a
s
ur
ing
we
ight
,
T
S
S
,
a
nd
pH
is
r
e
latively
s
tr
a
ight
f
or
wa
r
d,
but
a
s
s
e
s
s
ing
c
olor
c
ha
nge
pos
e
s
a
c
ons
ider
a
ble
c
ha
ll
e
nge
.
How
e
ve
r
,
pr
e
c
is
e
c
olo
r
c
h
a
nge
mea
s
ur
e
ment
is
c
r
uc
ial
f
o
r
e
va
luating
the
qu
a
li
ty
a
nd
s
he
lf
li
f
e
o
f
c
a
uli
f
lowe
r
f
lor
e
ts
,
of
f
e
r
ing
va
lua
ble
ins
ight
s
int
o
f
r
e
s
hne
s
s
a
nd
de
ter
ior
a
ti
on
ov
e
r
ti
me.
T
o
a
ddr
e
s
s
thi
s
c
ha
ll
e
nge
,
the
e
f
f
icie
nt
c
olor
c
ha
nge
mea
s
ur
e
ment
s
y
s
tem
is
int
r
oduc
e
d,
meticulous
ly
c
r
a
f
ted
to
a
c
c
ur
a
tely
qua
nti
f
y
c
olor
c
ha
nge
s
in
c
a
uli
f
lowe
r
f
lor
e
ts
.
T
he
pr
opos
e
d
c
olor
c
ha
nge
mea
s
u
r
e
ment
s
ys
tem
is
vis
ua
ll
y
de
picte
d
in
F
igu
r
e
3
.
At
the
c
or
e
of
thi
s
s
ys
tem,
it
incor
po
r
a
tes
r
e
d,
gr
e
e
n,
blue
(
R
GB
)
to
C
I
E
L
AB
c
olor
s
pa
c
e
c
onve
r
s
ion
to
e
nha
nc
e
c
olor
a
na
lys
is
.
C
onve
r
ti
ng
R
GB
c
ol
or
va
lues
to
the
C
I
E
L
AB
c
olor
s
pa
c
e
e
ns
ur
e
s
that
c
olor
f
e
a
tur
e
s
a
r
e
r
e
pr
e
s
e
nted
c
ons
is
tently
a
nd
pe
r
c
e
ptually
unif
or
m
,
im
pr
oving
the
a
c
c
ur
a
c
y
of
c
olor
c
ha
nge
mea
s
ur
e
ment.
Util
izing
the
c
onve
r
s
ion
e
qua
ti
ons
de
f
ined
by
the
I
nte
r
na
ti
ona
l
C
omm
is
s
ion
on
I
ll
u
mi
na
ti
on
(
C
I
E
)
,
the
R
GB
c
olor
va
lues
of
e
a
c
h
pixel
withi
n
the
s
e
gm
e
nted
r
e
gions
a
r
e
tr
a
ns
f
or
med
to
the
c
or
r
e
s
ponding
C
I
E
L
AB
c
olor
va
lues
.
T
he
c
onve
r
s
ion
e
qua
ti
ons
a
r
e
a
s
pr
e
s
e
nt
(
1)
-
(
3)
r
e
s
pe
c
ti
ve
ly,
whe
r
e
,
X,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Qualit
y
and
s
he
lf
-
li
fe
pr
e
diction
of
c
auli
fl
ow
e
r
us
i
ng
mac
hine
lear
ning
unde
r
v
ac
uum
…
(
M
d
A
pu
H
os
e
n
)
911
Y,
a
nd
Z
a
r
e
the
tr
is
ti
mul
us
va
lues
of
the
R
GB
c
olor
a
nd
X
n
,
Y
n
,
a
nd
Z
n
a
r
e
the
t
r
is
ti
mul
us
va
lu
e
s
of
the
r
e
f
e
r
e
nc
e
white
point
.
∗
=
116
×
(
)
−
16
(
1)
∗
=
500
×
[
(
)
−
(
)
]
(
2)
∗
=
200
×
[
(
)
−
(
)
]
(
3)
F
igur
e
3.
F
low
diagr
a
m
o
f
p
r
opos
e
d
c
olor
c
ha
nge
mea
s
ur
e
ment
s
ys
tem
Af
ter
that
the
s
ys
tem
li
e
s
the
int
e
gr
a
ti
on
of
a
dva
nc
e
d
im
a
ge
pr
oc
e
s
s
ing
tec
hn
iques
.
A
bil
a
ter
a
l
f
il
ter
is
a
ppli
e
d
dur
ing
pr
e
pr
oc
e
s
s
ing
to
r
e
f
ine
c
a
uli
f
low
e
r
f
lor
e
t
im
a
ge
s
.
T
his
f
il
ter
e
f
f
e
c
ti
ve
ly
r
e
duc
e
s
noi
s
e
while
pr
e
s
e
r
ving
c
r
it
ica
l
e
dge
s
a
nd
de
tails
,
e
ns
ur
ing
th
a
t
s
ubs
e
que
nt
c
olor
c
ha
nge
mea
s
ur
e
ment
is
c
onduc
ted
on
high
-
qua
li
ty
,
nois
e
-
f
r
e
e
im
a
ge
s
[
25]
.
F
oll
owing
p
r
e
pr
oc
e
s
s
ing,
pa
r
ti
c
le
s
wa
r
m
opti
mi
z
a
ti
on
(
P
S
O)
c
oupled
with
M
a
r
kov
r
a
ndom
f
ield
(
M
R
F
)
s
e
gmenta
ti
on
is
e
mpl
oye
d
to
opti
mi
z
e
the
s
e
gmenta
ti
on
pr
oc
e
s
s
.
T
he
pr
oc
e
s
s
be
gins
with
M
R
F
s
e
gmenta
ti
on,
whe
r
e
the
M
R
F
model
a
s
s
igns
labe
ls
to
pixels
ba
s
e
d
on
their
int
e
ns
it
ies
a
nd
s
pa
ti
a
l
r
e
lations
hips
,
e
ns
ur
ing
c
ohe
r
e
nt
a
nd
c
ontext
-
a
wa
r
e
s
e
gmenta
ti
on.
How
e
ve
r
,
the
e
f
f
e
c
ti
ve
ne
s
s
of
M
R
F
he
a
vil
y
de
pe
nds
on
the
c
h
oice
of
pa
r
a
mete
r
s
,
s
uc
h
a
s
the
we
ight
s
of
the
s
pa
ti
a
l
a
nd
int
e
ns
it
y
ter
ms
,
whic
h
a
r
e
opti
mi
z
e
d
us
ing
P
S
O
.
A
s
wa
r
m
of
pa
r
ti
c
l
e
s
r
e
pr
e
s
e
nts
potential
s
olut
ions
,
e
a
c
h
pa
r
ti
c
le
c
or
r
e
s
ponding
to
a
s
e
t
of
M
R
F
pa
r
a
mete
r
s
.
T
he
s
e
pa
r
ti
c
les
e
xplo
r
e
the
pa
r
a
mete
r
s
pa
c
e
,
upda
ti
ng
their
pos
it
ions
it
e
r
a
ti
ve
ly
ba
s
e
d
on
their
own
be
s
t
-
f
ound
s
olut
ion
a
nd
the
be
s
t
s
olut
ion
f
ound
by
th
e
s
wa
r
m.
Th
is
it
e
r
a
ti
ve
p
r
oc
e
s
s
c
onti
nue
s
unti
l
the
pa
r
ti
c
les
c
onve
r
ge
towa
r
ds
the
opti
mal
s
e
t
o
f
pa
r
a
mete
r
s
.
F
igur
e
4
s
hows
s
e
gmenta
ti
on
r
e
s
ult
s
of
L
,
a
,
b
c
ha
nne
l,
F
igur
e
4
(
a
)
p
r
e
s
e
nts
the
s
e
gmenta
ti
on
r
e
s
ult
s
us
ing
M
R
F
,
while
F
igur
e
4(
b
)
il
lus
tr
a
tes
the
opti
mi
z
e
d
M
R
F
s
e
gmenta
ti
on
outcome
s
.
B
y
c
ompar
ing
thes
e
f
igu
r
e
s
,
we
c
a
n
c
lea
r
ly
obs
e
r
ve
the
im
pr
ove
ments
int
r
odu
c
e
d
by
our
pr
opos
e
d
s
ys
tem,
h
ighl
ight
ing
it
s
e
f
f
e
c
ti
ve
ne
s
s
in
a
c
hieving
mor
e
a
c
c
ur
a
te
s
e
gmenta
ti
on
r
e
s
ult
s
.
F
inally,
to
a
s
s
e
s
s
the
tot
a
l
c
olo
r
c
ha
nge
a
c
r
os
s
t
he
c
a
uli
f
lowe
r
f
lo
r
e
ts
,
the
s
ys
tem
c
a
lcula
tes
the
de
lt
a
E
(
Δ
E
)
met
r
ic,
a
wide
ly
r
e
c
ognize
d
s
tanda
r
d
f
or
qua
nti
f
ying
c
olor
d
if
f
e
r
e
nc
e
s
a
c
r
os
s
dive
r
s
e
in
dus
tr
ies
,
including
f
ood
qua
li
ty
a
nd
pa
c
ka
ging.
T
his
metr
ic
a
c
c
ur
a
tely
c
a
ptur
e
s
the
pe
r
c
e
ptual
dif
f
e
r
e
nc
e
s
be
t
we
e
n
the
c
olor
s
of
im
a
ge
s
be
f
or
e
a
nd
a
f
ter
pa
c
ka
ging,
a
s
r
e
pr
e
s
e
nted
in
the
C
I
E
L
AB
c
olor
s
pa
c
e
.
B
y
e
va
luating
the
thr
e
e
c
omponents
(
L
,
a
,
a
nd
b)
,
Δ
E
pr
ov
ides
a
thor
ough
mea
s
ur
e
ment
o
f
c
olor
c
ha
nge
,
r
e
f
lec
ti
ng
a
ny
va
r
iations
due
to
the
e
f
f
e
c
ts
o
f
s
tor
a
ge
c
ondit
ions
ove
r
ti
me.
T
he
f
or
mul
a
f
or
Δ
E
is
de
tailed
in
(
4)
,
e
na
bli
ng
a
r
e
li
a
ble
a
nd
objec
ti
ve
a
s
s
e
s
s
ment
of
qua
li
ty
r
e
tenti
on
in
pa
c
ka
ge
d
c
a
uli
f
lowe
r
.
∆
=
√
(
∆
)
2
+
(
∆
)
2
+
(
∆
)
2
(
4)
I
n
(
4)
,
the
c
ha
nge
in
li
ghtnes
s
is
r
e
pr
e
s
e
nted
by
(
∆
L
)
,
whi
le
(
∆
a
)
,
a
nd
(
∆
b)
,
c
or
r
e
s
pond
to
s
hif
t
s
in
the
gr
e
e
n
-
r
e
d
a
nd
blue
-
ye
ll
ow
c
olor
c
ha
nne
l
s
,
r
e
s
pe
c
ti
ve
ly.
B
y
int
e
gr
a
ti
ng
thes
e
thr
e
e
c
omponents
,
the
Δ
E
metr
ic
pr
ovides
a
r
e
l
iable
a
nd
holi
s
ti
c
mea
s
ur
e
o
f
th
e
ove
r
a
ll
c
olor
c
ha
nge
.
T
his
a
ll
ows
us
to
de
t
e
c
t
e
ve
n
s
ubtl
e
s
hif
ts
in
c
olor
,
whic
h
a
r
e
c
r
uc
ial
f
or
a
s
s
e
s
s
ing
the
f
r
e
s
hne
s
s
a
nd
qua
li
ty
of
the
c
a
uli
f
lowe
r
f
lor
e
ts
thr
oughout
s
tor
a
ge
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
2,
Apr
il
2025
:
907
-
9
16
912
(
a
)
(
b)
F
igur
e
4.
S
e
gmenta
ti
on
r
e
s
ult
s
of
L
,
a
,
b
c
ha
nne
l
u
s
ing
(
a
)
M
R
F
a
nd
(
b)
opti
mi
z
e
d
M
R
F
s
e
gmenta
ti
on
3.
3.
Dat
as
e
t
p
r
e
p
ar
at
ion
B
e
f
or
e
c
onduc
ti
ng
the
e
xpe
r
im
e
nts
,
meticulous
p
r
e
pa
r
a
ti
on
of
the
da
tas
e
t
wa
s
e
s
s
e
nti
a
l
to
e
ns
ur
e
the
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bil
it
y
of
the
r
e
s
ult
s
.
T
he
da
tas
e
t
c
ons
is
ted
of
mea
s
ur
e
ments
obtaine
d
f
r
om
va
r
i
ous
tes
ts
c
onduc
ted
on
c
a
uli
f
lowe
r
s
a
mpl
e
s
ove
r
a
s
e
r
ies
o
f
pe
r
iods
pos
t
-
pa
c
ka
ging.
S
pe
c
if
ica
ll
y,
the
da
tas
e
t
i
nc
luded
mea
s
ur
e
ments
of
we
ight
los
s
,
pH
,
T
S
S
,
a
nd
c
olo
r
c
ha
nge
f
or
e
a
c
h
c
a
uli
f
lowe
r
s
a
mpl
e
a
t
r
e
gular
i
nter
va
ls
.
T
o
f
a
c
il
it
a
te
uni
f
or
mi
ty
a
nd
c
ons
is
tenc
y
in
da
t
a
c
oll
e
c
ti
on,
s
tanda
r
dize
d
pr
otocols
we
r
e
f
o
ll
o
we
d
f
or
c
onduc
ti
ng
e
a
c
h
tes
t,
including
s
a
mpl
e
pr
e
pa
r
a
ti
o
n,
mea
s
ur
e
ment
pr
oc
e
dur
e
s
,
a
nd
r
e
c
or
ding
metho
dologi
e
s
.
S
tr
ingent
qua
li
ty
c
ontr
ol
mea
s
ur
e
s
we
r
e
im
pleme
nted
to
mi
ti
ga
te
potential
s
our
c
e
s
of
e
r
r
o
r
a
nd
e
n
s
ur
e
the
int
e
gr
it
y
of
the
da
tas
e
t.
3.
4
.
F
e
a
t
u
r
e
s
e
lec
t
ion
T
o
a
s
s
e
s
s
the
s
ign
if
ica
nc
e
of
va
r
ious
pa
r
a
mete
r
s
,
i
nc
lud
ing
T
S
S
,
c
o
lor
c
ha
ng
e
,
we
ig
ht
los
s
,
a
nd
p
H,
a
na
lys
is
of
v
a
r
i
a
nc
e
(
A
NO
VA
)
wa
s
e
m
plo
ye
d
.
A
N
OV
A
is
a
s
tat
is
ti
c
a
l
tec
hn
ique
t
ha
t
de
t
e
r
m
i
ne
s
whe
t
he
r
the
r
e
a
r
e
s
ta
ti
s
t
ica
ll
y
s
ig
ni
f
ica
n
t
d
if
f
e
r
e
nc
e
s
be
twe
e
n
th
e
mea
ns
of
th
r
e
e
or
mo
r
e
g
r
o
ups
[
26
]
.
T
he
r
e
s
ul
t
s
of
the
AN
OV
A
tes
ts
a
r
e
s
um
mar
ize
d
in
T
a
b
le
1,
w
he
r
e
both
th
e
F
-
va
lue
a
n
d
p
-
va
lue
we
r
e
c
ons
ide
r
e
d
i
n
s
e
lec
ti
n
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Qualit
y
and
s
he
lf
-
li
fe
pr
e
diction
of
c
auli
fl
ow
e
r
us
i
ng
mac
hine
lear
ning
unde
r
v
ac
uum
…
(
M
d
A
pu
H
os
e
n
)
913
ke
y
pr
e
dict
or
s
.
Va
r
iab
les
w
it
h
p
-
va
lues
un
de
r
0
.
0
5
we
r
e
de
e
me
d
s
tat
is
ti
c
a
ll
y
s
ig
ni
f
ica
n
t,
a
nd
hi
ghe
r
F
-
va
lues
indi
c
a
ted
a
s
tr
onge
r
va
r
ianc
e
a
m
ong
g
r
oup
mea
ns
,
unde
r
s
c
or
i
ng
the
i
mpo
r
ta
nc
e
of
t
he
r
e
s
pe
c
ti
ve
pa
r
a
mete
r
.
T
he
f
ind
ings
in
T
a
ble
1
r
e
ve
a
l
that
T
S
S
,
c
olor
c
ha
nge
,
a
nd
we
ight
los
s
a
r
e
s
tatis
ti
c
a
ll
y
s
igni
f
ica
nt
pr
e
dictor
s
of
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
he
lf
li
f
e
,
a
s
indi
c
a
ted
by
thei
r
p
-
va
lues
f
a
ll
ing
unde
r
the
s
ign
if
ica
nc
e
thr
e
s
hold
of
0
.
05.
F
u
r
ther
mor
e
,
their
r
e
latively
hi
gh
F
-
va
lues
18.
63
,
22.
48
,
a
nd
20
.
36,
r
e
s
pe
c
ti
ve
ly
highl
ight
the
s
ubs
tantial
dif
f
e
r
e
nc
e
s
in
gr
oup
mea
ns
,
r
e
i
nf
or
c
ing
the
r
e
leva
nc
e
of
thes
e
f
e
a
tur
e
s
f
or
pr
e
dictive
modeling.
C
ons
e
que
ntl
y,
T
S
S
,
c
olo
r
c
ha
nge
,
a
nd
we
ight
los
s
we
r
e
s
e
lec
ted
a
s
f
e
a
tur
e
s
f
or
de
ve
lo
ping
the
pr
e
dictive
models
.
I
n
c
ontr
a
s
t,
pH
did
not
mee
t
the
c
r
it
e
r
ia
f
or
s
tatis
ti
c
a
l
s
igni
f
ica
nc
e
,
with
a
p
-
va
lue
of
0
.
349
a
nd
a
low
F
-
va
lue
of
1.
46,
s
ugge
s
ti
ng
mi
nim
a
l
va
r
ianc
e
a
c
r
os
s
gr
oup
mea
n
s
.
T
his
r
e
s
ult
indi
c
a
tes
that
pH
c
ontr
ibut
e
s
mi
nim
a
ll
y
to
f
or
e
c
a
s
ti
ng
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
he
lf
li
f
e
in
th
is
s
tudy,
lea
ding
to
it
s
e
xc
lus
ion
f
r
om
the
f
inal
pr
e
dictive
model
.
T
a
ble
1
.
R
e
s
ult
s
of
AN
OV
A
f
or
f
e
a
tur
e
s
e
lec
ti
on
V
a
r
ia
bl
e
F
-
va
lu
e
p
-
va
lu
e
T
S
S
18.63
0.025
C
ol
or
C
ha
nge
22.48
0.009
W
e
ig
ht
L
os
s
20.36
0.017
pH
1.46
0.349
3.
5
.
M
ac
h
in
e
lear
n
in
g
m
o
d
e
li
n
g
a
n
d
m
od
e
l
e
val
u
at
ion
T
o
pr
e
dict
the
qua
li
ty
a
nd
s
he
lf
li
f
e
,
it
is
ne
c
e
s
s
a
r
y
to
know
the
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
on
a
s
pe
c
if
ic
da
y.
M
a
c
hine
lea
r
ning
tec
hniques
a
r
e
e
mpl
oye
d
to
pr
e
dict
the
we
ight
los
s
,
c
olo
r
c
ha
nge
,
a
nd
T
S
S
o
f
c
a
uli
f
lowe
r
a
s
de
pe
nde
nt
va
r
ia
bles
,
with
the
number
of
da
ys
pos
t
-
pa
c
ka
ging
a
s
the
inde
pe
nde
nt
va
r
iable
.
T
his
a
ppr
oa
c
h
a
ll
ows
f
o
r
t
r
a
c
king
the
dy
na
mi
c
c
ha
nge
s
in
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
ove
r
ti
me.
T
he
p
r
e
dicte
d
we
ight
los
s
,
c
olor
c
ha
nge
,
a
n
d
T
S
S
a
r
e
c
ompar
e
d
with
a
thr
e
s
hold
va
lue,
a
nd
ba
s
e
d
on
thi
s
,
the
qua
li
ty
a
nd
s
he
lf
li
f
e
o
f
c
a
uli
f
lowe
r
unde
r
s
pe
c
if
ic
pa
c
ka
ging
c
ondit
ions
a
r
e
pr
e
dicte
d.
F
or
p
r
e
diction,
two
mac
hine
lea
r
ning
a
lgo
r
it
hms
,
AN
N
a
nd
li
ne
a
r
r
e
gr
e
s
s
ion,
a
r
e
e
mpl
oye
d.
L
inea
r
r
e
gr
e
s
s
ion
models
a
r
e
uti
li
z
e
d
to
e
s
tabl
is
h
r
e
lations
hips
be
twe
e
n
the
indepe
nde
nt
va
r
iable
(
da
ys
pos
t
-
pa
c
k
a
ging)
a
nd
the
de
pe
nde
nt
va
r
iable
s
(
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
)
.
Additi
ona
ll
y
,
AN
N
is
us
e
d
to
c
a
ptur
e
nonli
ne
a
r
de
pe
nde
nc
ies
a
nd
int
r
ica
te
pa
tt
e
r
ns
withi
n
the
da
ta,
a
im
ing
to
e
nha
nc
e
the
a
c
c
ur
a
c
y
of
pr
e
dictions
.
T
he
pe
r
f
or
manc
e
o
f
thes
e
two
a
lg
or
it
hms
is
e
va
luate
d
us
ing
the
R
-
s
qua
r
e
d
va
lue,
p
r
ovidi
ng
va
luable
ins
ight
s
int
o
the
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bil
it
y
of
the
models
in
pr
e
dicting
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
ove
r
ti
me.
B
y
c
ompa
r
ing
the
r
e
s
ult
s
of
th
e
s
e
two
a
ppr
oa
c
he
s
,
ins
ight
s
a
r
e
de
ve
loped
int
o
whic
h
a
lgor
it
hm
pe
r
f
o
r
ms
be
tt
e
r
in
f
o
r
e
c
a
s
ti
ng
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
he
lf
li
f
e
,
ther
e
by
in
f
or
mi
ng
de
c
is
ion
-
making
in
f
ood
pr
oduc
ti
on
a
nd
dis
tr
ibut
io
n
pr
oc
e
s
s
e
s
.
3.
6
.
Qu
al
it
y
a
n
d
s
h
e
lf
-
li
f
e
p
r
e
d
ict
ion
Us
ing
the
pr
e
dicte
d
va
lues
f
or
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
,
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
he
lf
l
if
e
a
r
e
e
s
ti
mate
d
by
c
ompar
ing
thes
e
va
lues
with
e
s
tablis
he
d
qua
li
ty
th
r
e
s
holds
.
T
he
s
e
th
r
e
s
holds
a
c
t
a
s
e
s
s
e
nti
a
l
be
nc
hmar
ks
,
of
f
e
r
ing
ins
ight
s
int
o
the
c
a
uli
f
lowe
r
'
s
qua
li
ty
a
nd
e
xpe
c
ted
longevity
unde
r
s
pe
c
if
ic
pa
c
ka
ging
c
ondit
ions
.
T
he
e
xpe
r
im
e
nt
e
s
tablis
he
d
that
a
m
a
xim
um
a
c
c
e
ptable
tot
a
l
c
olor
c
ha
nge
of
a
ppr
o
xim
a
tely
3
unit
s
is
s
e
t
f
or
c
a
uli
f
lowe
r
mea
nt
f
or
mar
ke
t;
e
xc
e
e
ding
thi
s
thr
e
s
hold
s
ignals
a
noti
c
e
a
ble
de
c
li
ne
in
vis
ua
l
a
ppe
a
l
a
nd
ove
r
a
ll
qua
li
ty,
making
the
c
a
uli
f
lowe
r
uns
uit
a
ble
f
or
s
a
le.
S
im
il
a
r
ly
,
a
we
ight
los
s
unde
r
15
%
is
c
ons
ider
e
d
e
s
s
e
nti
a
l
to
maintain
the
c
a
uli
f
lowe
r
’
s
f
r
e
s
hne
s
s
a
nd
textur
e
,
a
s
e
xc
e
s
s
ive
we
ight
los
s
i
nd
ica
tes
de
hydr
a
ti
on
a
nd
s
igni
f
ica
nt
moi
s
tur
e
los
s
.
M
or
e
ove
r
,
T
S
S
leve
ls
up
to
7
.
2
a
r
e
r
e
ga
r
de
d
a
s
id
e
a
l
f
or
c
a
uli
f
lowe
r
,
e
nha
nc
ing
it
s
tas
te
a
nd
ove
r
a
ll
c
ons
umer
a
ppe
a
l.
Highe
r
T
S
S
leve
ls
ge
ne
r
a
ll
y
c
or
r
e
late
with
be
tt
e
r
qua
li
ty
,
whic
h
is
c
r
uc
ial
f
or
both
co
ns
umer
s
a
ti
s
f
a
c
ti
on
a
nd
mar
ke
t
va
lue.
B
y
a
li
gni
ng
the
pr
e
dicte
d
we
ight
los
s
,
c
olo
r
c
ha
nge
,
a
nd
T
S
S
va
lues
with
thes
e
p
r
e
de
f
ined
thr
e
s
holds
,
a
c
c
ur
a
te
f
o
r
e
c
a
s
ts
a
bout
the
qua
li
ty
a
nd
r
e
maining
s
he
lf
li
f
e
of
c
a
u
li
f
lowe
r
c
a
n
be
made
.
C
a
uli
f
lowe
r
mee
ti
ng
the
th
r
e
s
hold
c
r
it
e
r
ia
is
c
las
s
if
ied
a
s
high
qua
li
ty
with
the
potent
ial
f
o
r
a
longer
s
he
lf
li
f
e
.
C
onve
r
s
e
ly,
whe
n
a
ny
of
thes
e
li
mi
ts
a
r
e
e
xc
e
e
de
d,
the
c
a
uli
f
lowe
r
's
qua
li
ty
is
li
ke
ly
to
de
ter
ior
a
te
mo
r
e
r
a
pidl
y,
r
e
quir
ing
a
s
hor
ter
s
he
lf
-
li
f
e
e
s
ti
mate
.
T
his
a
pp
r
oa
c
h
e
ns
ur
e
s
that
only
c
a
uli
f
lowe
r
of
opti
mal
qua
li
ty
r
e
a
c
he
s
the
mar
ke
t,
a
li
gn
ing
wit
h
qua
li
ty
a
s
s
ur
a
nc
e
s
tanda
r
ds
.
4.
RE
S
UL
T
AN
D
DI
S
CU
S
S
I
ON
T
his
s
tudy
inves
ti
ga
ted
the
pr
e
diction
o
f
c
a
uli
f
lo
we
r
qua
li
ty
a
nd
s
he
lf
li
f
e
us
ing
mac
hine
lea
r
n
ing
unde
r
va
c
uum
a
nd
M
AP.
W
hil
e
pr
e
vious
r
e
s
e
a
r
c
h
ha
s
e
xplor
e
d
the
im
pa
c
t
o
f
va
r
ious
pa
c
ka
ging
methods
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
2,
Apr
il
2025
:
907
-
9
16
914
pe
r
is
ha
ble
goods
s
uc
h
a
s
f
r
uit
s
a
nd
f
is
h,
it
ha
s
not
thor
oughly
a
ddr
e
s
s
e
d
the
inf
luenc
e
of
c
os
t
-
e
f
f
e
c
ti
ve
va
c
uum
a
nd
c
he
mi
c
a
l
-
b
a
s
e
d
M
AP
on
c
a
uli
f
lowe
r
.
T
his
ga
p
is
s
ign
if
ica
nt
due
to
the
high
pe
r
is
ha
bi
li
ty
a
nd
e
c
onomi
c
im
por
tanc
e
of
c
a
uli
f
lowe
r
.
Our
f
indi
ngs
r
e
ve
a
l
that
in
c
he
mi
c
a
l
-
ba
s
e
d
M
AP
a
nd
va
c
uum
pa
c
ka
ging
s
ys
tems
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
ha
ve
a
gr
e
a
t
im
pa
c
t
on
qua
li
ty
a
nd
s
he
lf
.
T
he
pr
opos
e
d
method
s
igni
f
ica
ntl
y
e
nha
nc
e
s
the
a
c
c
ur
a
c
y
of
p
r
e
dicting
ke
y
qua
li
ty
pa
r
a
mete
r
s
,
incl
uding
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
.
T
he
no
ve
l
c
olor
c
ha
nge
mea
s
ur
e
ment
s
ys
tem,
whic
h
int
e
gr
a
tes
bil
a
ter
a
l
f
il
ter
ing
with
P
S
O
a
nd
M
R
F
s
e
gmenta
ti
on,
a
c
hieve
d
a
n
int
e
r
s
e
c
ti
on
ove
r
union
(
I
oU
)
s
c
or
e
o
f
0
.
96
pla
ys
a
gr
e
a
t
r
ole
in
the
pr
e
diction
of
ke
y
qua
li
ty
pa
r
a
mete
r
s
.
T
his
pe
r
f
or
manc
e
s
ur
pa
s
s
e
s
that
of
a
lt
e
r
na
ti
ve
s
e
gmenta
ti
on
tec
hniques
,
a
s
s
hown
in
T
a
ble
2,
whic
h
c
ompar
e
s
the
I
oU
s
c
or
e
s
of
d
if
f
e
r
e
nt
methods
.
T
a
ble
2
.
C
ompar
is
on
of
s
e
gmenta
ti
on
methods
(
I
o
U
s
c
or
e
s
)
S
e
gme
nt
a
ti
on
me
th
od
I
oU
s
c
or
e
P
r
opos
e
d
m
e
th
od
0.96
K
-
me
a
ns
c
lu
s
te
r
in
g
0.87
W
a
te
r
s
he
d t
r
a
ns
f
or
ma
ti
on
0.81
R
e
gi
on gr
ow
in
g
0.92
T
he
I
oU
s
c
or
e
f
o
r
the
pr
opos
e
d
method
is
not
a
bly
higher
than
thos
e
o
f
K
-
mea
ns
c
lus
ter
ing,
wa
ter
s
he
d
tr
a
ns
f
or
mation,
a
nd
r
e
gion
gr
owing
,
d
e
mons
tr
a
ti
ng
s
upe
r
ior
a
c
c
ur
a
c
y
in
de
li
ne
a
ti
ng
c
a
uli
f
lowe
r
f
lor
e
ts
a
nd
mea
s
ur
ing
s
ubtl
e
c
olo
r
c
ha
nge
s
.
A
hig
he
r
I
oU
s
c
or
e
indi
c
a
tes
be
tt
e
r
s
e
gmenta
ti
on
a
c
c
ur
a
c
y
a
nd
c
ons
is
tenc
y
[
27]
,
s
ugge
s
ti
ng
that
the
pr
opos
e
d
m
e
thod
a
c
hieve
s
a
mo
r
e
pr
e
c
is
e
s
e
gmenta
ti
on
o
f
in
divi
dua
l
f
lor
e
ts
,
lea
ding
to
im
pr
ove
d
mea
s
ur
e
ment
o
f
c
olor
c
ha
nge
s
.
Us
ing
thi
s
pr
opos
e
d
c
olor
c
ha
nge
mea
s
ur
e
ment
method
in
c
onjunction
with
mac
hine
lea
r
ning
modeling,
the
AN
N
models
de
mons
tr
a
ted
high
pr
e
dictive
a
c
c
ur
a
c
y
a
c
r
os
s
va
r
ious
pa
c
ka
ging
methods
,
with
R
-
s
qu
a
r
e
d
va
lues
of
0.
981
f
or
c
olor
c
ha
nge
.
T
he
AN
N
a
ls
o
a
c
quir
e
d
the
R
s
qua
r
e
d
va
lue
0.
99
2
f
or
we
ight
lo
s
s
,
a
nd
0.
952
f
o
r
T
S
S
.
T
he
s
e
r
e
s
ult
s
,
de
tailed
in
T
a
ble
3,
c
ompar
e
the
R
-
s
qua
r
e
d
va
lues
f
or
li
ne
a
r
r
e
gr
e
s
s
ion
a
nd
AN
N
models
unde
r
di
f
f
e
r
e
nt
pa
c
ka
ging
c
ondit
ions
,
f
ur
ther
h
ighl
ight
ing
the
r
obus
tnes
s
of
the
AN
N
a
pp
r
oa
c
h
in
pr
e
dicting
qua
li
ty
metr
ics
f
o
r
c
a
uli
f
lowe
r
.
T
a
ble
3
.
P
r
e
diction
r
e
s
ult
of
dif
f
e
r
e
nt
pa
r
a
mete
r
s
f
or
dif
f
e
r
e
nt
pa
c
ka
ging
P
a
r
a
me
te
r
s
P
a
c
ka
gi
ng
L
in
e
a
r
r
e
gr
e
s
s
io
n
ANN
C
ol
or
C
ha
nge
M
A
P
1
0.926
0.978
M
A
P
2
0.912
0.970
M
A
P
3
0.931
0.981
V
a
c
uum
0.940
0.962
W
e
ig
ht
l
os
s
M
A
P
1
0.944
0.983
M
A
P
2
0.961
0.990
M
A
P
3
0.962
0.985
V
a
c
uum
0.982
0.992
T
S
S
M
A
P
1
0.893
0.911
M
A
P
2
0.886
0.934
M
A
P
3
0.904
0.946
V
a
c
uum
0.923
0.952
T
he
s
tudy's
r
e
s
ult
s
s
how
that
AN
N
models
c
ons
is
tently
outper
f
or
m
li
ne
a
r
r
e
gr
e
s
s
ion
models
in
pr
e
dicting
c
a
uli
f
lowe
r
qua
li
ty
pa
r
a
mete
r
s
.
F
or
c
olor
c
ha
nge
,
the
AN
N
model
a
c
hieve
d
R
-
s
qua
r
e
d
va
lues
r
a
nging
f
r
om
0.
96
to
0.
98
a
c
r
os
s
dif
f
e
r
e
nt
pa
c
ka
ging
methods
.
F
or
we
ight
los
s
,
the
AN
N
model
r
e
a
c
he
d
R
-
s
qu
a
r
e
d
va
lues
of
up
to
0.
99
,
a
nd
f
or
T
S
S
,
the
R
-
s
qua
r
e
d
va
lues
we
r
e
be
twe
e
n
0.
91
a
nd
0
.
95.
T
o
c
ontextua
li
z
e
thes
e
r
e
s
ult
s
,
the
r
e
s
ult
is
c
ompar
e
d
with
f
indi
ngs
f
r
om
other
r
e
leva
nt
s
tudi
e
s
,
a
s
s
umm
a
r
ize
d
in
T
a
ble
4
.
C
ompar
ing
the
s
tudy's
r
e
s
ult
s
wi
th
p
r
e
v
ious
r
e
s
e
a
r
c
h
s
hows
that
the
a
ppr
oa
c
h
with
mor
e
pr
e
c
is
e
c
olor
c
ha
nge
mea
s
ur
e
ment
us
ing
AN
N
models
a
c
h
ieve
d
higher
a
c
c
ur
a
c
y.
T
a
ble
4
.
C
ompar
is
on
of
c
a
uli
f
lowe
r
qua
li
ty
a
nd
s
h
e
lf
-
li
f
e
pr
e
diction
methods
S
tu
dy
P
a
c
ka
g
in
g
m
e
th
od
R
-
s
qua
r
e
d
v
a
lu
e
s
C
ur
r
e
nt
s
tu
dy
M
A
P
(
C
he
mi
c
a
l
ba
s
e
d
)
a
nd V
a
c
uum
0.992
M
oha
mm
e
d
e
t
al
.
[
20]
M
A
P
(
G
a
s
mi
xt
ur
e
)
a
nd V
a
c
uum
0.951
A
ld
e
n
e
t
al
.
[
23]
M
A
P
(
G
a
s
mi
xt
ur
e
)
0.990
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
Ar
ti
f
I
ntell
I
S
S
N:
2252
-
8938
Qualit
y
and
s
he
lf
-
li
fe
pr
e
diction
of
c
auli
fl
ow
e
r
us
i
ng
mac
hine
lear
ning
unde
r
v
ac
uum
…
(
M
d
A
pu
H
os
e
n
)
915
How
e
ve
r
,
the
methods
a
nd
r
e
s
ult
s
may
not
ge
n
e
r
a
li
z
e
a
c
r
os
s
a
ll
c
a
uli
f
lowe
r
va
r
ieties
o
r
other
pe
r
is
ha
ble
ve
ge
table
s
.
F
ur
ther
r
e
s
e
a
r
c
h
is
ne
e
de
d
to
c
onf
i
r
m
the
a
ppli
c
a
bil
it
y
of
thes
e
tec
hniques
in
va
r
ious
r
e
a
l
-
wor
ld
c
ondit
ions
,
s
uc
h
a
s
dif
f
e
r
ing
s
tor
a
ge
a
nd
tr
a
ns
por
tation
e
nvir
onme
nts
.
F
utur
e
s
tudi
e
s
c
ould
f
oc
us
on
a
da
pti
ng
the
tec
hnology
f
o
r
c
omm
e
r
c
ial
us
e
on
a
lar
ge
r
s
c
a
le
a
nd
opt
im
izing
the
mac
hine
-
lea
r
ning
models
by
incor
por
a
ti
ng
diver
s
e
da
ta
s
e
ts
a
nd
e
nvir
onment
a
l
va
r
iable
s
.
5.
CONC
L
USI
ON
T
hr
ough
r
igor
ous
e
xpe
r
im
e
ntation
a
nd
a
na
lys
is
,
the
s
igni
f
ica
nc
e
of
f
e
a
tur
e
s
s
uc
h
a
s
we
ight
los
s
,
c
olor
c
ha
nge
,
a
nd
T
S
S
in
pr
e
dicting
c
a
uli
f
lowe
r
q
ua
li
ty
ove
r
ti
me
in
c
he
mi
c
a
l
(
M
AP)
a
nd
va
c
uum
p
a
c
ka
ging
ha
s
be
e
n
de
mons
tr
a
ted.
T
he
nove
l
c
olo
r
c
ha
nge
mea
s
ur
e
m
e
nt
s
ys
tem,
int
e
gr
a
ti
ng
a
dva
nc
e
d
s
e
gmenta
ti
on
tec
hniques
a
nd
c
olor
s
pa
c
e
c
onve
r
s
ion,
pr
ovides
a
n
a
c
c
ur
a
te
a
nd
e
f
f
icie
nt
a
s
s
e
s
s
ment
of
c
olor
c
h
a
nge
s
in
c
a
uli
f
lowe
r
f
lor
e
ts
.
Additi
ona
ll
y
,
mac
hine
lea
r
nin
g
a
lgor
it
hms
,
including
li
ne
a
r
r
e
gr
e
s
s
ion
a
nd
AN
N
,
e
xhi
bit
pr
omi
s
ing
pe
r
f
or
manc
e
in
pr
e
dicting
ke
y
qua
li
ty
i
ndica
tor
s
.
T
he
s
e
f
indi
ngs
ha
ve
im
pli
c
a
ti
ons
f
or
opti
mi
z
ing
pa
c
ka
ging
s
tr
a
tegie
s
,
mi
nim
izing
f
ood
wa
s
te,
a
nd
e
ns
ur
ing
the
de
li
ve
r
y
o
f
high
-
qua
li
ty
c
a
uli
f
l
owe
r
to
c
ons
umer
s
.
F
utur
e
r
e
s
e
a
r
c
h
c
ould
f
oc
us
on
r
e
f
in
ing
pr
e
dictive
models
,
e
xplor
ing
a
ddit
ional
f
e
a
tu
r
e
s
,
a
nd
va
li
da
ti
ng
r
e
s
ult
s
a
c
r
os
s
dif
f
e
r
e
nt
s
tor
a
ge
c
ondit
ion
s
a
nd
c
a
uli
f
lowe
r
va
r
ieties
.
AC
KNOWL
E
DGE
M
E
NT
S
W
e
e
xtend
our
s
ince
r
e
g
r
a
ti
tude
to
the
I
C
T
Divis
ion,
M
ini
s
tr
y
o
f
P
os
ts
,
T
e
lec
omm
unica
ti
ons
,
a
nd
I
nf
or
mation
T
e
c
hnology
,
B
a
nglade
s
h,
f
or
their
invalua
ble
s
uppor
t
in
f
und
ing
thi
s
r
e
s
e
a
r
c
h
und
e
r
gr
a
nt
number
56.
00
.
0000.
052
.
33.
001
.
23
-
61.
T
his
a
s
s
is
tanc
e
ha
s
be
e
n
ins
tr
umenta
l
in
a
dva
nc
ing
ou
r
wor
k,
a
nd
we
a
r
e
de
e
ply
a
ppr
e
c
iative
of
their
c
omm
it
ment
to
f
os
ter
ing
innovation
a
nd
de
ve
lopm
e
nt.
RE
F
E
RE
NC
E
S
[
1]
S
.
A
kt
he
r
,
M
.
R
.
I
s
la
m,
M
.
A
la
m,
M
.
J
.
A
la
m,
a
nd
S
.
A
hme
d,
“
I
mpa
c
t
of
s
li
ght
ly
a
c
id
ic
e
le
c
tr
ol
yz
e
d
w
a
te
r
in
c
ombi
na
ti
on
w
it
h
ul
tr
a
s
ound
a
nd
mi
ld
he
a
t
on
s
a
f
e
ty
a
nd
qua
li
ty
of
f
r
e
s
h
c
ut
c
a
ul
if
lo
w
e
r
,”
P
os
th
ar
v
e
s
t
B
io
lo
gy
and
T
e
c
hnol
ogy
,
vol
.
197,
M
a
r
.
2023,
doi
:
10.1016/j
.pos
th
a
r
vbi
o.2022.112189.
[
2]
J
.
J
.
M
in
i
e
t
al
.
,
“
I
nve
s
ti
ga
ti
on
o
f
a
nt
im
ic
r
obi
a
l
a
nd
a
nt
i
-
c
a
nc
e
r
a
c
ti
vi
ty
of
th
e
r
ma
ll
y
s
e
ns
it
iv
e
S
n
O
2
na
nos
tr
uc
tu
r
e
s
w
it
h
gr
e
e
n
-
s
ynt
he
s
iz
e
d
c
a
ul
if
lo
w
e
r
mor
phol
ogy
a
t
a
mbi
e
nt
w
e
a
th
e
r
c
o
ndi
ti
ons
,”
E
nv
ir
onm
e
nt
al
R
e
s
e
ar
c
h
,
vol
.
245,
M
a
r
.
2024,
doi
:
10.1016/j
.e
nvr
e
s
.2023.117878.
[
3]
T
.
A
.
A
.
N
a
s
r
in
e
t
al
.
,
“
P
r
e
s
e
r
va
ti
on
of
pos
th
a
r
ve
s
t
qua
li
ty
of
f
r
e
s
h
c
ut
c
a
u
li
f
lo
w
e
r
th
r
ough
s
im
pl
e
a
nd
e
a
s
y
pa
c
ka
gi
ng
te
c
hni
que
s
,”
A
ppl
ie
d F
ood R
e
s
e
ar
c
h
, vol
. 2, no. 2, De
c
. 2022, d
oi
:
10.1016/j
.a
f
r
e
s
.2022.100125.
[
4]
F
.
J
our
a
bi
a
n
a
nd
M
.
N
our
i,
“
O
pt
im
iz
a
ti
on
of
f
or
mul
a
te
d
ke
f
ir
a
n/
ma
lv
a
ne
gl
e
c
ta
f
il
m
w
it
h
r
ic
e
br
a
n
oi
l
to
ma
in
ta
in
c
a
ul
if
lo
w
e
r
qua
li
ty
in
s
to
r
a
ge
,”
P
r
oc
e
e
di
ngs
of
th
e
N
at
io
nal
A
c
ade
m
y
of
Sc
ie
nc
e
s
,
I
ndi
a
Se
c
ti
on
B
:
B
io
lo
gi
c
al
Sc
ie
nc
e
s
,
vol
.
93,
no.
3,
pp. 697
–
703, S
e
p. 2023, doi:
10.1007/s
40011
-
023
-
01463
-
6.
[
5]
D
.
S
ha
r
ma
,
M
.
J
.
A
la
m,
I
.
A
.
B
e
gum,
S
.
D
in
g,
a
nd
A
.
M
.
M
c
K
e
nz
ie
,
“
A
va
lu
e
c
ha
in
a
na
ly
s
is
of
c
a
ul
if
lo
w
e
r
a
nd
to
ma
to
in
B
a
ngl
a
de
s
h
,
”
Sus
ta
in
abi
li
ty
, vol
. 15, no. 14, J
ul
. 2023, doi:
10.
3390/s
u151411395.
[
6]
G
. W
u
e
t
al
.
, “
R
e
gul
a
ti
on of
r
e
s
pi
r
a
to
r
y r
a
te
a
nd s
to
r
a
ge
qua
li
ty
of
pos
th
a
r
ve
s
t
c
a
ul
if
lo
w
e
r
ba
s
e
d on ga
s
pe
r
me
a
bi
li
ty
modi
f
ic
a
ti
on
us
in
g
ga
s
ba
r
r
ie
r
(
GB
)
-
ga
s
c
onduc
to
r
(
GC
)
b
le
ndi
ng
pa
c
ka
gi
ng,”
F
ood
P
ac
k
agi
ng
and
She
lf
L
if
e
,
vol
.
39,
N
ov.
2023,
doi
:
10.1016/j
.f
ps
l.
2023.101161.
[
7]
K
.
J
a
dw
is
ie
ńc
z
a
k,
Z
.
K
a
li
ni
e
w
ic
z
,
S
.
K
onopka
,
D
.
C
hos
z
c
z
,
a
nd
J
.
M
a
jk
ow
s
ka
-
G
a
do
ms
ka
,
“
A
pr
opos
a
l
f
or
a
pr
oc
e
s
s
in
g
li
ne
f
or
c
a
ul
if
lo
w
e
r
a
nd br
oc
c
ol
i
f
lo
r
e
tt
in
g,”
A
ppl
ie
d Sc
ie
nc
e
s
, vol
. 13, no. 4, F
e
b. 2023, doi:
10.3390/app130425
09.
[
8]
Z
.
W
a
ng,
Q
.
L
i,
S
.
J
ia
ng,
X
.
W
a
ng,
S
.
W
a
ng,
a
nd
L
.
H
ou,
“
I
mpr
ovi
ng
r
a
di
o
f
r
e
que
nc
y
he
a
ti
ng
uni
f
or
mi
ty
in
c
a
ul
if
lo
w
e
r
by
c
ha
ngi
ng
de
ns
it
y
in
di
f
f
e
r
e
nt
z
one
s
,”
F
ood
and
B
io
pr
oduc
ts
P
r
oc
e
s
s
in
g
,
vol
.
143,
pp.
1
–
8,
J
a
n.
2024,
doi
:
10.1016/j
.f
bp.2023.10.004.
[
9]
Q
.
J
ia
ng,
M
.
Z
ha
ng,
A
.
S
.
M
uj
umda
r
,
a
nd
B
.
C
he
n,
“
C
o
mpa
r
a
ti
ve
f
r
e
e
z
in
g
s
tu
dy
of
br
oc
c
ol
i
a
nd
c
a
ul
i
f
lo
w
e
r
:
e
f
f
e
c
ts
of
e
le
c
tr
os
ta
ti
c
f
ie
ld
a
nd s
ta
ti
c
ma
gn
e
ti
c
f
ie
ld
,”
F
ood
C
he
m
is
t
r
y
, v
ol
. 397, De
c
. 2022, doi:
10.1016/j
.f
oodc
he
m.2022.133751.
[
10]
K
.
K
a
yna
ş
,
“
I
gl
o
ka
r
na
ba
ha
r
ç
e
ş
id
in
in
nor
ma
l
ve
kont
r
ol
lu
a
tm
os
f
e
r
koş
ul
la
r
ın
da
de
pol
a
nma
s
ın
,”
J
ou
r
nal
of
A
gr
i
c
ul
tu
r
al
F
ac
ul
ty
of
G
az
io
s
m
anpas
a U
ni
v
e
r
s
it
y
, vol
. 37, no. 2020
–
2, pp. 94
–
101,
J
a
n. 2020, doi:
10.13002/j
a
f
a
g4609.
[
11]
L
.
W
a
ng
a
nd
M
.
T
e
pl
it
s
ki
,
“
M
i
c
r
obi
ol
ogi
c
a
l
f
ood
s
a
f
e
ty
c
on
s
id
e
r
a
ti
ons
in
s
he
lf
-
li
f
e
e
xt
e
ns
io
n
of
f
r
e
s
h
f
r
ui
ts
a
nd
ve
g
e
ta
bl
e
s
,”
C
ur
r
e
nt
O
pi
ni
on i
n B
io
te
c
hnol
ogy
, vol
. 80, Apr
. 2023, doi:
10.
1016/j
.c
opbi
o.2023.102895.
[
12]
B
.
N
a
vi
na
,
K
.
K
.
H
ut
ha
a
s
h,
N
.
K
.
V
e
lm
ur
uga
n,
a
nd
T
.
K
or
umi
ll
i,
“
I
ns
ig
ht
s
in
to
r
e
c
e
nt
in
nova
ti
ons
in
a
nt
i
br
ow
ni
ng
s
tr
a
te
gi
e
s
f
or
f
r
ui
t
a
nd ve
ge
ta
bl
e
pr
e
s
e
r
va
ti
on,”
T
r
e
nd
s
in
F
ood Sc
ie
nc
e
& T
e
c
hnol
ogy
, vol
. 139, S
e
p. 2023, doi:
10.1016/j
.t
if
s
.2023.104128.
[
13]
V
.
M
a
r
ti
ns
,
M
.
P
in
ta
do,
R
.
M
or
a
is
,
a
nd
A
.
M
or
a
is
,
“
R
e
c
e
nt
hi
ghl
ig
ht
s
in
s
u
s
ta
in
a
bl
e
bi
o
-
ba
s
e
d
e
di
bl
e
f
il
ms
a
nd
c
o
a
ti
ngs
f
or
f
r
ui
t
a
nd ve
ge
ta
bl
e
a
ppl
ic
a
ti
ons
,”
F
oods
,
vol
. 13, no. 2, J
a
n. 2024, d
oi
:
10.3390/f
oods
13020318.
[
14]
J
.
A
lv
e
s
,
P
.
D
.
G
a
s
pa
r
,
T
.
M
.
L
im
a
,
a
nd
P
.
D
.
S
il
va
,
“
W
ha
t
is
th
e
r
ol
e
of
a
c
ti
ve
pa
c
ka
gi
ng
in
th
e
f
ut
ur
e
o
f
f
ood
s
us
ta
in
a
bi
li
ty
?
a
s
ys
te
ma
ti
c
r
e
vi
e
w
,
”
J
our
nal
of
th
e
Sc
ie
nc
e
of
F
ood
and
A
gr
ic
ul
tu
r
e
,
vol
.
103,
no.
3,
pp.
1004
–
1020,
F
e
b.
2023,
doi
:
10.1002/j
s
f
a
.11880.
[
15]
M
. M
ul
la
n a
nd D
. M
c
D
ow
e
ll
, “
M
odi
f
ie
d a
tm
os
phe
r
e
pa
c
ka
gi
n
g,”
i
n
F
ood
and B
e
v
e
r
age
P
ac
k
agi
ng T
e
c
hnol
ogy
, W
il
e
y, 2011,
pp.
263
–
294
, doi
:
10.1002/978144439
2180.c
h10.
[
16]
K
.
W
.
M
c
M
il
li
n,
“
M
odi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng,
”
in
F
ood
E
ngi
ne
e
r
in
g
Se
r
ie
s
,
S
pr
in
ge
r
,
C
ha
m,
2020,
pp.
693
–
718
,
doi
:
10.1007/978
-
3
-
030
-
42660
-
6_26.
[
17]
M
.
S
iv
e
r
ts
vi
k,
J
.
T
.
R
os
ne
s
,
a
nd
H
.
B
e
r
gs
li
e
n,
“
M
odi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng,”
in
M
in
imal
pr
oc
e
s
s
in
g
te
c
hnol
ogi
e
s
in
th
e
fo
od
in
dus
tr
y
, C
a
mbr
id
ge
:
W
oodh
e
a
d P
ubl
is
hi
ng L
td
, 2002, pp. 61
–
86.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
Ar
ti
f
I
ntell
,
Vol.
14,
No.
2,
Apr
il
2025
:
907
-
9
16
916
[
18]
P
.
Q
u,
M
.
Z
ha
ng,
K
.
F
a
n,
a
nd
Z
.
G
uo,
“
M
ic
r
opor
ous
modi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng
to
e
xt
e
nd
s
he
lf
li
f
e
of
f
r
e
s
h
f
oods
:
a
r
e
vi
e
w
,”
C
r
it
ic
al
R
e
v
ie
w
s
in
F
ood
Sc
ie
nc
e
and
N
ut
r
it
io
n
,
v
ol
.
62,
no.
1,
pp.
51
–
65,
J
a
n.
2022,
doi
:
10.1080/10408398.2
020.1811635.
[
19]
A
.
A
.
K
a
de
r
,
D
.
Z
a
gor
y,
E
.
L
.
K
e
r
be
l,
a
nd
C
.
Y
.
W
a
ng,
“
M
odi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng
of
f
r
ui
ts
a
nd
ve
g
e
ta
bl
e
s
,”
C
r
it
i
c
al
R
e
v
ie
w
s
i
n F
ood Sc
ie
nc
e
and
N
ut
r
it
io
n
, vol
. 28, no. 1, pp
. 1
–
30, J
a
n. 1989, doi:
10.1080/104083989
09527490.
[
20]
M
.
M
oha
mm
e
d,
R
.
S
r
in
iv
a
s
a
g
a
n,
A
.
A
lz
a
hr
a
ni
,
a
nd
N
.
K
.
A
lq
a
ht
a
ni
,
“
M
a
c
hi
ne
-
le
a
r
ni
ng
-
ba
s
e
d
s
pe
c
tr
os
c
opi
c
te
c
hni
que
f
or
non
-
de
s
tr
uc
ti
ve
e
s
ti
ma
ti
on
of
s
he
lf
li
f
e
a
nd
qua
li
ty
of
f
r
e
s
h
f
r
ui
ts
pa
c
ka
ge
d
unde
r
modi
f
ie
d
a
tm
os
phe
r
e
s
,”
Sus
ta
in
abi
li
ty
,
vol
.
15,
no. 17, Aug. 2023, d
oi
:
10.3390/s
u151712871.
[
21]
D
.
A
lb
e
r
t
-
W
e
is
s
a
nd
A
.
O
s
ma
n,
“
I
nt
e
r
a
c
ti
ve
de
e
p
le
a
r
ni
ng
f
or
s
he
lf
li
f
e
pr
e
di
c
ti
on
of
mus
kme
lo
ns
ba
s
e
d
on
a
n
a
c
ti
ve
l
e
a
r
ni
ng
a
ppr
oa
c
h,”
Se
ns
o
r
s
, vol
. 22, no
. 2,
J
a
n. 2022, doi:
10.3390/s
220
20414.
[
22]
I
. B
. I
or
li
a
m, B
. A
.
I
kyo, A.
I
or
li
a
m,
E
. O
. O
kube
, K
. D
. K
w
a
g
ht
yo, a
nd Y
. I
. S
he
hu, “
A
ppl
ic
a
ti
on of
ma
c
hi
ne
l
e
a
r
ni
ng t
e
c
hni
q
ue
s
f
or
okr
a
s
he
lf
li
f
e
pr
e
di
c
ti
on,”
J
our
nal
of
D
at
a
A
nal
y
s
is
and
I
nf
or
m
at
io
n
P
r
oc
e
s
s
i
ng
,
vol
.
9,
no.
3,
pp.
136
–
150,
2021,
doi
:
10.4236/j
da
ip
.2021.93009.
[
23]
K
.
M
.
A
ld
e
n,
M
.
O
mi
d,
A
.
R
a
ja
bi
pour
,
B
.
T
a
je
ddi
n,
a
nd
M
.
S
.
F
ir
ouz
,
“
Q
ua
li
ty
a
nd
s
he
lf
-
li
f
e
pr
e
di
c
ti
on
of
c
a
ul
if
lo
w
e
r
un
de
r
modi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng
by
us
in
g
a
r
ti
f
ic
ia
l
ne
ur
a
l
n
e
twor
ks
a
nd
im
a
ge
pr
oc
e
s
s
in
g,”
C
om
put
e
r
s
and
E
le
c
tr
oni
c
s
in
A
gr
ic
ul
tu
r
e
, vol
. 163, Aug. 2019, do
i:
10.1016/j
.c
ompa
g.2019.
104861.
[
24]
Z
.
F
u,
S
.
Z
ha
o,
X
.
Z
ha
ng,
M
.
P
ol
ovka
,
a
nd
X
.
W
a
ng,
“
Q
ua
li
ty
c
ha
r
a
c
te
r
is
ti
c
s
a
na
ly
s
is
a
nd
r
e
ma
in
in
g
s
h
e
lf
li
f
e
pr
e
di
c
ti
on
of
f
r
e
s
h
ti
be
ta
n
tr
ic
hol
oma
ma
ts
ut
a
ke
unde
r
modi
f
ie
d
a
tm
os
phe
r
e
pa
c
ka
gi
ng
in
c
ol
d
c
ha
in
,”
F
oods
,
vol
.
8,
no.
4,
A
pr
.
2019,
doi
:
10.3390
/f
oods
8040136.
[
25]
S
.
P
a
r
is
,
P
.
K
or
npr
obs
t,
J
.
T
umbl
in
,
a
nd
F
.
D
ur
a
nd,
B
il
at
e
r
a
l
fi
lt
e
r
in
g:
th
e
or
y
and
appl
ic
at
io
n
s
,
F
ound
a
ti
ons
a
nd
T
r
e
nd
s
®
in
C
omput
er
G
r
a
ph
ic
s
a
nd
V
is
io
n
, vol
. 4, no. 1, pp. 1
–
75, 2008, doi:
10.1561/060000002
0.
[
26]
R
.
N
.
H
e
ns
on,
“
A
na
ly
s
i
s
of
va
r
ia
nc
e
(
ANOVA
)
,”
in
B
r
ai
n
M
appi
ng
,
E
ls
e
vi
e
r
,
2015,
pp.
477
–
481
,
doi
:
10.1016/B
978
-
0
-
12
-
397025
-
1.00319
-
5.
[
27]
F
.
V.
B
e
e
r
s
,
A
.
L
in
ds
tr
öm,
E
.
O
ka
f
or
,
a
nd
M
.
W
ie
r
in
g,
“
D
e
e
p
ne
ur
a
l
ne
twor
ks
w
it
h
in
te
r
s
e
c
ti
on
ove
r
uni
on
lo
s
s
f
or
bi
na
r
y
im
a
ge
s
e
gme
nt
a
ti
on,”
in
P
r
oc
e
e
di
ngs
of
th
e
8t
h
I
nt
e
r
nat
io
nal
C
onf
e
r
e
nc
e
on
P
at
te
r
n
R
e
c
ogni
ti
on
A
ppl
ic
at
io
ns
and
M
e
th
ods
,
S
C
I
T
E
P
R
E
S
S
-
S
c
ie
nc
e
a
nd T
e
c
hnol
ogy P
ubl
ic
a
ti
ons
, 2019, pp
. 438
–
445
, doi
:
10.5220/000734750
4380445.
B
I
OG
RA
P
HI
E
S
OF
AU
T
HO
RS
Md
.
A
pu
Ho
s
en
i
s
a
mas
t
er's
s
t
u
d
e
n
t
p
u
rs
u
i
n
g
Co
m
p
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eeri
n
g
a
t
J
as
h
o
re
U
n
i
v
er
s
i
t
y
o
f
Sc
i
en
ce
an
d
T
ech
n
o
l
o
g
y
,
Ban
g
l
a
d
es
h
.
H
e
h
o
l
d
s
a
b
ach
e
l
o
r's
d
e
g
ree
i
n
Co
mp
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
ee
ri
n
g
fro
m
t
h
e
s
ame
u
n
i
v
ers
i
t
y
.
H
i
s
res
earch
i
n
t
ere
s
t
s
en
co
mp
a
s
s
mach
i
n
e
l
ear
n
i
n
g
,
d
eep
l
earn
i
n
g
,
co
mp
u
t
er
v
i
s
i
o
n
,
e
x
p
l
a
i
n
a
b
l
e
A
I,
an
d
n
e
t
w
o
rk
i
n
g
.
H
e
can
b
e
c
o
n
t
act
e
d
at
emai
l
:
a
p
u
.
c
s
e.
j
u
s
t
@
g
ma
i
l
.
c
o
m.
D
r.
Sy
ed
M
d.
Ga
l
i
b
i
s
a
p
ro
fe
s
s
o
r
i
n
t
h
e
d
e
p
art
me
n
t
o
f
Co
m
p
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eeri
n
g
(CSE
)
at
J
as
h
o
re
U
n
i
v
er
s
i
t
y
o
f
Sci
en
ce
an
d
T
ech
n
o
l
o
g
y
(J
U
ST
),
Ban
g
l
ad
e
s
h
.
Befo
re
j
o
i
n
i
n
g
at
J
U
S
T
,
h
e
w
as
w
o
rk
i
n
g
as
an
A
s
s
o
ci
at
e
Pro
fe
s
s
o
r
i
n
t
h
e
D
ep
ar
t
men
t
o
f
Co
mp
u
t
er
Sci
e
n
ce
a
n
d
In
f
o
rma
t
i
o
n
T
ech
n
o
l
o
g
y
a
t
Pat
u
a
k
h
a
l
i
Sci
en
ce
a
n
d
T
ech
n
o
l
o
g
y
U
n
i
v
er
s
i
t
y
,
Ban
g
l
a
d
es
h
.
H
e
fi
n
i
s
h
e
d
h
i
s
P
h
.
D
.
fr
o
m
Sw
i
n
b
u
r
n
e
U
n
i
v
er
s
i
t
y
o
f
T
ech
n
o
l
o
g
y
,
A
u
s
t
ra
l
i
a
i
n
2
0
1
5
.
Befo
re
t
h
at
,
h
e
co
mp
l
e
t
ed
h
i
s
M.
Sc.
i
n
Co
mp
u
t
er
E
n
g
i
n
eeri
n
g
fro
m
D
al
ar
n
a
U
n
i
v
er
s
i
t
y
,
Sw
ed
e
n
i
n
2
0
0
8
.
H
e
o
b
t
a
i
n
e
d
h
i
s
B.
Sc.
i
n
Co
mp
u
t
er
Sci
e
n
ce
an
d
E
n
g
i
n
eeri
n
g
fro
m
K
h
u
l
n
a
U
n
i
v
ers
i
t
y
,
Ban
g
l
ad
e
s
h
i
n
2
0
0
5
.
Cu
rren
t
l
y
,
he
i
s
t
h
e
ch
a
i
rman
o
f
t
h
e
D
ep
ar
t
men
t
o
f
CSE
,
J
U
ST
.
Mo
reo
v
er,
h
e
l
ead
s
a
res
earch
g
r
o
u
p
n
amed
art
i
fi
c
i
al
i
n
t
e
l
l
i
g
e
n
ce
fo
r
d
ev
e
l
o
p
men
t
(A
ID
g
r
o
u
p
)
w
i
t
h
i
n
t
h
e
D
ep
ar
t
men
t
o
f
CSE
.
H
i
s
res
earc
h
i
n
t
eres
t
i
s
mo
s
t
l
y
i
n
t
h
e
fi
el
d
o
f
i
mag
e
p
ro
ce
s
s
i
n
g
an
d
ar
t
i
f
i
ci
a
l
i
n
t
el
l
i
g
en
ce
a
p
p
l
i
e
d
t
o
s
o
ci
a
l
d
e
v
el
o
p
me
n
t
.
H
e
can
b
e
co
n
t
act
e
d
at
emai
l
:
g
a
l
i
b
.
cs
e@
j
u
s
t
.
ed
u
.
b
d
.
Evaluation Warning : The document was created with Spire.PDF for Python.