TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.7, July 201
4, pp
. 5414 ~ 54
1
9
DOI: 10.115
9
1
/telkomni
ka.
v
12i7.537
1
5414
Re
cei
v
ed
De
cem
ber 1
2
, 2013; Re
vi
sed
F
ebruary 23,
2014; Accept
ed March 1
5
, 2014
The Navel Orange Sugar and Acidity Quantitative
Prediction Model Optimization Research by Second
Generation Wavelet Transform
Zhao Ke*, Yang Han, Wan
g
Zhong, Wa
ng Qi
Coll
eg
e of Information En
gi
ne
erin
g, Nanch
a
n
g
Han
g
ko
ng U
n
iversit
y
, Jia
n
g
xi, Na
n Cha
n
g
,
China
*Corres
p
o
n
id
n
g
author, em
ail
:
zhaoke6
80
5
@
12
6.com
A
b
st
r
a
ct
T
he auth
o
r res
earch
es the i
m
pact of the sec
ond
ge
nerati
o
n
w
a
velet transf
o
rm s
pectro
m
e
t
er dat
a
prepr
ocessi
ng
nave
l
or
an
ge s
ugar
conte
n
t a
nd
acid
ity Pa
rti
a
l L
east S
q
u
a
r
e
s (PLS)
qu
ant
itative
accurac
y
o
f
the pr
edicti
o
n
mod
e
l. T
h
is
pap
er a
l
so c
o
llects th
e sp
e
c
tral d
a
te of
one
hu
ndr
ed
nave
l
or
an
ges
b
y
visibl
e/ne
ar-infr
a
red
diffuse r
e
flectanc
e d
e
tection
tec
h
n
o
l
ogy a
nd
esta
blish
e
s the
na
vel or
ang
e su
ga
r
content a
nd ac
idity PLS pr
edi
ction
mo
d
e
l us
i
ng the sixty n
a
v
el ora
n
g
e
s as
the establ
ishi
n
g
sa
mpl
e
s. T
h
e
author
co
ntrasts cha
n
g
e
s of
nave
l
or
an
ge
sugar
co
nt
ent
and
aci
d
ity P
L
S pre
d
ictio
n
mode
l b
e
ca
use
th
e
spectral
date
of nave
l
or
an
ges ar
e pr
oce
ssed by
the
secon
d
g
ener
ation w
a
v
e
let
transform, F
i
n
a
l
l
y
concl
u
sio
n
: the
secon
d
g
e
n
e
r
a
tion w
a
v
e
let t
r
ansfor
m
pr
oce
ssing
nav
el
orang
e sp
ectral
data c
an i
m
pro
v
e
the pred
ictive a
b
ility of the
su
g
a
r content a
nd
acidity PLS
q
u
antitative a
n
a
l
y
s
is mo
de
ls.
Ke
y
w
ords
:
near
infrar
ed
spectru
m
, sec
ond
g
ener
atio
n w
a
ve
l
e
t
tran
sform, partia
l
least
s
q
u
a
re, nave
l
oran
ge, aci
d
ity, sugar conte
n
t
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Navel
ora
n
g
e
taste
s
goo
d, full of jui
c
e with
out se
ed,
go
od qu
ality,
bright color,
a
nd
contai
ns all kinds of nutri
e
n
ts,whi
ch a
r
e
human
nee
d
.
In addition to focusi
ng o
n
the external
quality of navel ora
nge, for example si
ze
, color
a
nd
sh
ape, co
nsum
ers
pay more
attention to the
sug
a
r, a
c
idity
,
vitamin cont
ent and
taste
,
which
a
r
e t
he inte
rnal
q
uality indicators . Suga
r a
nd
acidity is an
important in
dicato
r of na
vel orang
e, sug
a
r an
d a
c
idity largely
depen
ds on
the
amount
and
types of fruit su
gar an
d o
r
gani
c
aci
d
s,
etc. So the
rese
arch
of th
e navel
oran
ge
sug
a
r a
nd a
c
i
d
ity by the internal
quality of rapi
d, a
c
cu
rate an
d non
-destructive d
e
tection m
e
th
od
has far-rea
chi
ng sig
n
ifica
n
ce.
With the developme
n
t of comp
uter an
d el
ectroni
c sci
en
ce and
techn
o
logy a
nd the
physi
cal a
nd
chemi
c
al
met
r
ology, n
ear-i
nfrared
spe
c
t
r
osco
py tech
nology d
e
vel
opment i
s
ra
pid
in re
ce
nt
y
e
a
r
s,
it
ha
s m
o
re a
d
v
ant
ag
e
s
,
s
u
ch
a
s
:
q
u
ic
k an
aly
s
i
s
spe
ed
qui
ckl
y
,
less
sam
p
le
pretreatment,
gree
n pollu
tion-fre
e
and
non-de
stru
ctive, so it has be
co
me
a hot re
sea
r
ch.
Dome
stic an
d
foreig
n rese
arche
r
s
have
done
a lot of
resea
r
ch work in th
e a
nalyzed i
n
ternal fruit
quality indi
cat
o
rs. M
oghi
mi
(201
0) [1] u
s
ed ne
ar-i
nfra
red spe
c
tro
scopy fore
ca
st the SSC a
nd
PH
of ki
wi, analy
z
ed
the e
s
tab
lish m
odel
predictio
n's
effe
ct by u
s
ing
dif
f
erent p
r
etrea
t
ment metho
d
s
;
He Yong (20
06) [2] use
d
325nm
-1
075
nm near-infra
r
ed spe
c
tro
s
copy to predict
the quality of the
different bay
berry base on method
s
of prin
ci
pal
comp
one
nt analysi
s
and
neural netwo
rk.
Niep
eng
Ch
eng, Yang
Yan (2
010
)
[3] used
pri
n
cip
a
l comp
onent a
nalysis an
d comb
ines
Bayesian li
ne
ar di
scrimin
a
n
t and ne
cta
r
forwa
r
d
neu
ral network predictio
n mod
e
l, comp
arative
analysi
s
of th
e accu
ra
cy of
pre
d
iction
m
odel
s, t
he co
nclu
sio
n
is
u
s
ing vi
sible
a
nd ne
ar i
n
fra
r
e
d
techn
o
logy fo
r h
oney fa
st
cla
ssifi
cation
is fea
s
ibl
e
; Li
u Yan
de
(20
12) [4] u
s
ed
reverse
inte
rval
partial l
e
a
s
t
squ
a
re
s met
hod, g
eneti
c
algorith
m
a
n
d
continu
o
u
s
proje
c
tion
al
gorithm,
built
th
e
near-infrared
spe
c
tro
s
copy
partial lea
s
t squ
a
re
s regression mo
del
of apple sol
u
ble soli
ds.
The expe
rim
ental coll
ect
spe
c
tral
data
of the navel oran
ge thro
ugh diffuse reflectan
c
e
near-infrared
spe
c
tro
s
copy
detection te
chnolo
g
y, bec
ause the out
put sign
al of spe
c
tral d
a
ta
in
the colle
ction
terminal is
affected by h
a
rmo
n
ics or
each harmon
i
c, voltage waveform will
be
mixed with m
o
re inte
rfere
n
c
e an
d noise, it will
appear glitche
s
spi
k
e, it's phase also
will have
a
larger m
o
vement, in order to i
m
prove the
ac
curacy and
reli
ability
of prediction models,
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
The Na
vel O
r
ange Sug
a
r a
nd Acidity
Qu
antitative Pre
d
iction Mo
del
Optim
i
zation… (Zha
o Ke)
5415
comp
re
hen
si
vely analysi
s
of dom
esti
c and
fore
i
g
n
schol
ars o
n
fruit spe
c
tra
l
pre
p
rocessi
n
g
method
s. Thi
s
de
sign u
s
es the seco
nd gen
erat
io
n wavelet transfo
rm for t
he sp
ectral data
prep
ro
ce
ssin
g of navel orange, an
d then create
o
r
ange’
s suga
r and aci
d
ity conte
n
t of pa
rtial
least squ
a
res
(PLS) qua
ntitative
predi
cti
on
mo
del
[5
-7], and
com
p
arative a
naly
s
is the
cha
n
g
e
s
of perfo
rman
ce b
e
twe
en
before
and
a
fter the PLS
predi
ction
m
odel u
s
in
g seco
nd g
ene
ration
wavelet algorithm .
2. Materials
and Method
s of Mea
s
ure
m
ent
2.1. Material's Selection a
nd Classi
fic
a
tion
Experimental
sam
p
le
s co
ntained
100
Gan
z
ho
u na
vel ora
nge
were p
u
rcha
sed f
r
o
m
sup
e
rm
arket i
n
Na
nchang,
then culled th
e sam
p
le
s, which
ero
ded
b
y
pests, d
a
m
aged
or
sha
p
e
d
oddly. After collecte
d
nave
l
oran
ge sam
p
les,tu
rned th
em numb
e
r
randomly, an
d
then pla
c
ed
in
a co
nsta
nt te
mperature
ch
ambe
r with
2
5
Ԩ
in 1
2
h
o
urs, th
e day
that the exp
e
rime
ntal data
measured m
u
st b
e
same
with th
e d
a
y
of navel
o
r
ang
e colle
ct
ed.
Spe
c
ific requi
rem
ents
of
c
l
as
s
i
fic
a
tion
c
a
tegories
as s
h
own in the following table:
Table 1. Mea
s
uri
ng Cl
assif
i
cation of Sa
mples
group1 calibratio
n
set
group 2 suga
r pr
ediction set
group 3 acidit
y
p
r
ediction
set
sample selection
method
number 1
-
60
number 61
-80
number 81
-100
number of sampl
e
s
60
20
20
collected spectra
equatorial parts
of uniform thr
ee-
point
equatorial parts
of uniform
three-p
o
int
equatorial parts
of uniform
three-p
o
int
detection navel
orange
sugar and acidit
y througho
ut the
navel orange
sugar of the
whole orange
acidity
of th
e
w
h
ole orange
2.2. Selectio
n of Spectro
meter Expe
riment
and the Method of
Spectral Acquisition
Experiment
s used LS
-1
type tung
st
en h
a
loge
n
light sou
r
ce, TCD13
0
4
AP linear
detecto
rs, po
rtable sp
ectrometer USB
4000
to
coll
e
c
t the
spe
c
tral data of
na
vel ora
nge
s,
the
oran
ge
spe
c
t
r
um'
s
data
a
nalyse
d
thro
ugh t
he
ove
r
ture
software to and
se
t the scanni
ng
con
d
ition
s
of sampl
e
s: inte
gration time i
s
50m
s,
sm
o
o
th width i
s
two pixel
s
, the band
sele
cte
d
in
the ra
nge
o
f
400-140
0n
m. By near-infra
red
sp
e
c
tro
s
copy dif
f
use
refle
c
ta
nce
dete
c
tio
n
techni
que, co
llected 1
0
poi
nts sp
ectral data fr
om the
three mea
s
u
r
eme
n
t point
s on equ
atori
a
l
parts for the
study of
100
oran
ge ,
an
d
then ave
r
ag
e
d
30
the
spe
c
tral
data
of t
he
sam
p
les,
the
each numb
e
r
is the ea
ch n
a
vel oran
ge'
s spe
c
tral data
.
2.3. Sugar and Acidit
y
Ph
y
s
icochemical Metho
d
s
of Mea
s
ure
m
ent
Usi
ng a h
and
held glu
c
o
s
e
meter
WYT-4
type in acco
rdan
ce
with t
he nation
a
l st
anda
rd
GB1229
5-90
to measu
r
e t
he su
gar of g
r
oup 1 a
nd g
r
oup 3 sample
s. Usi
ng ha
n
dheld PH m
e
te
r
PHSJ-4A type according t
o
the national
standa
rd
GB
/T 5009 1-20
03 GB/T 124
56-9
0
to mea
s
ure
the acidity of orang
e from
group 1 a
n
d
grou
p 3 sample
s. Sug
a
r and a
c
idit
y measu
r
em
ent
requi
rem
ents for continuo
us mea
s
u
r
e
m
ent 6 times as a wh
ol
e averag
ed sug
a
r an
d acidity
values.
2.4. Spectral
Data Pre
p
ro
cessing M
e
thods
Becau
s
e
in
a
ddition to th
e
origi
nal
spe
c
trum
sign
al containe
d info
rmation
relate
d to the
chemi
c
al
structure of th
e materi
al, b
u
t also
cont
ained
sig
nal
noise g
ene
rated from
m
any
confo
undi
ng factors, the
wavelet tran
sfo
r
m can
sh
o
w
the nature of
non-stat
ion
a
ry signal
s in ti
me
domain
and
freque
ncy d
o
m
ain bette
r, thus removin
g
the the
s
e i
n
terferen
ce o
f
signal
s u
s
e
d
wavelet tran
sform. Con
s
tru
c
tor
of secon
d
gen
er
ation
wavelet tran
sform al
gorith
m
enh
an
ce
s t
h
e
gene
ration of
wavelet tran
sform, the sp
eed of wavel
e
t algorithm
enabl
es fa
ste
r
. The ba
sic i
dea
of the method is con
s
truct
ed by
the co
nventional wavelet filter, decom
po
sed t
he basi
c
buil
d
ing
blocks, com
p
leted the wa
velet transfo
rmation
by st
eppin
g
se
qu
entially, it can be summa
rized
as: re
solutio
n
,
predictio
n a
nd upd
ating.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5414 – 54
19
5416
(1) Decomp
o
s
ition ca
n be inert wavel
e
t transfo
rm, the origin
al sig
nal
()
Xn
decom
po
sed
into odd an
d even numb
e
r,
the sample
were set a
s
follows:
a
n=
2
n
,
XX
n
Z
(1)
b
n=
2
n
+
1
,
XX
n
Z
(2)
(2) F
o
recast:
there i
s
a
certain co
rrelatio
n
between th
e origi
nal
sa
mples
of ea
ch sig
nal,
the coeffici
en
t
a
n
X
used to pred
ict
b
n
X
,the predi
ction op
erato
r
P generate
d
n
c
n
d
+
+
+
+
+
+
b
n
X
n
X
a
n
X
b
n
X
n
X
a
n
X
, namely:
nb
a
d=
n
n
XP
X
(3)
(3) Up
date: In
ord
e
r to
re
du
ce the
alia
sin
g
effect
s of th
e wavel
e
t tra
n
sform an
d p
r
eserve
the origin
al
a
n
X
chara
c
te
risti
c
s of certain freque
nci
e
s, u
s
ed a
n
ope
rator set A to produ
ce
better data
n
c
,
it can be expressed a
s
:
na
n
c=
n
+
d
XA
(4)
Secon
d
gen
eration
wav
e
let recon
s
truction p
r
o
c
e
ss i
s
the reverse pro
c
ess of
decompo
sitio
n
, that is a
n
ti-upd
ate a
nd predi
ction
and d
e
com
positio
n. Re
constructio
n
a
nd
decompo
sitio
n
of expressi
on is th
e sa
me, just
n
e
e
d
to ch
ang
e
the sig
n
an
d
orde
r. As
sh
own
belo
w
schem
atic blo
ck di
a
g
ram of de
co
mpositio
n an
d recon
s
tru
c
ti
on:
Figure 1. Second Ge
neration Wavel
e
t Reco
nstruc
tio
n
and Sche
ma
tic Diag
ram o
f
the Principle
of Decompo
s
i
t
ion
2.5. Predictiv
e
Modeling and Ev
aluation of Param
e
ter
By using
st
oichi
o
metry
softwa
r
e
un
scra
mble
r8.0
and m
a
tlab2
007 tool
s fo
r data
pro
c
e
ssi
ng, t
he expe
rime
nt usin
g a li
near mod
e
l
of partial l
e
a
s
t squa
re
s (PLS) metho
d
to
establi
s
h a q
uantitative predictio
n mod
e
l for the su
g
a
r and
acidity
of navel ora
nge. Rei
n
stit
uted
the PLS qua
ntitative pred
iction m
odel
of sug
a
r
and
acidity cont
ent of orang
e usi
ng
se
co
nd
gene
ration
wavelet transfo
rm pre
p
ro
ce
ssing spe
c
tr
o
s
copy data, Compa
r
ative analysi
s
usin
g the
se
con
d
gen
eration wavel
e
t transfo
rm of
spe
c
tr
al
data
prep
ro
ce
ssin
g before and
after the nav
el
oran
ge
su
ga
r an
d a
c
idit
y PLS quan
titative predi
ction m
odel
perfo
rma
n
ce chang
es.
The
perfo
rman
ce
of the m
odel
evaluated
by
mod
e
ling
th
e correlatio
n
coeffici
ent, m
odelin
g the
root
mean
squa
re
erro
r
,
mod
e
l
predi
ction
correlation
co
efficient a
n
d
ro
ot mea
n
squ
a
re
e
rro
r of
model p
r
edi
ction. The hig
her
co
rrel
a
tio
n
co
e
fficient
model, mod
e
ling the smal
ler root mea
n
squ
a
re e
r
ror
and the ro
ot mean squa
re
predi
ction e
r
ro
r, the stro
ng
er the ability to predi
ct mod
e
l.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
The Na
vel O
r
ange Sug
a
r a
nd Acidity
Qu
antitative Pre
d
iction Mo
del
Optim
i
zation… (Zha
o Ke)
5417
3. Experimental Re
sults
and An
aly
s
is
3.1. The Effe
ct o
f
Nav
e
l
Orang
e
Sug
a
r PLS Mode
l b
y
Second
Gener
a
tion
Wav
e
let Spe
c
tral
Data Prepro
cessing
Acco
rdi
ng to the method
mentione
d in
sectio
n 1.2 acq
u
isitio
n correctio
n
and
2 set of
predi
ction
sets of n
a
vel o
r
ange
sp
ect
r
a
l
data,
in
se
ction 1.3
stan
dard
phy
sical
and
ch
emical
measurement
method for d
e
termini
ng co
rre
ction an
d
2 set of predi
ction set
s
act
ual sug
a
r val
ue
of
the navel oran
ge,
e
s
ta
blish
a
qua
ntitative pre
d
icti
on m
odel
PL
S of the
su
g
a
r
co
ntent of
the
oran
ge, then
use
d
the pre
d
iction m
odel
to predi
ct
20
navel ora
nge
's suga
r of 2 grou
p; usin
g the
se
con
d
ge
ne
ration
wavele
t transfo
rm f
o
r n
a
vel o
r
a
nge
spe
c
tral
data p
r
ep
ro
cessing, to
bu
ild
navel ora
nge
suga
r PLS q
uantitative predictio
n mod
e
l, then use
d
20 navel ora
nge'
s of 2 group
to pre
d
ict the
sug
a
r. As
sh
own i
n
Figu
re
2, Figure
3, i
t
is the di
scre
te points
dist
ribution of
sug
a
r
navel o
r
an
ge
perce
ntage
of p
r
edi
cted
and
a
c
tual
before
an
d
after the
second
gen
erati
o
n
wavelet transform:
Analysis two cha
r
ts sho
w
s that the origi
nal ora
nge b
r
ix PLS model and optimiza
t
ion of
se
con
d
ge
ne
ration
wavele
t correlatio
n
coeffi
ci
ent orange
brix PL
S model, sta
ndard deviati
on
corre
c
tion
of
the mo
del, th
e p
r
edi
ction
model
of
roo
t
mean
squa
re
error valu
es
as
sho
w
n
in
table:
Figure 2. The
Brix PLS Model of Origin
a
l
Navel Orang
e
Figure 3. The
Brix PLS Model of Navel
Oran
ge
after
Seco
nd Gene
ration Wavelet
Pretreatm
ent
Analysis two cha
r
ts sho
w
s that the origi
nal ora
nge b
r
ix PLS model and optimiza
t
ion of
se
con
d
ge
ne
ration
wavele
t correlatio
n
coeffi
ci
ent orange
brix PL
S model, sta
ndard deviati
on
corre
c
tion
of
the mo
del, th
e p
r
edi
ction
model
of
roo
t
mean
squa
re
error valu
es
as
sho
w
n
in
table:
Table 2. The
Analysis
Re
sults of the Ori
g
inal an
d Sse
c
on
d Gen
e
rat
i
on Wavel
e
t Optimizatio
n
Navel Orang
e Brix Partial Lea
st
Square
s
(PLS) Mm
ethod Mod
e
l
Number
The
corr
elation
coefficient of
model
Correc
t
i
on
standard
deviation of model
Standard deviati
on of prediction
model
PLS model of ori
g
inal Brix
20
90.34
0.763
0.806
PLS model of op
timization
Brix
20 94.56
0.696
0.725
It can
be
se
en that
the
spectral
data
with n
a
vel o
r
ange
after th
e second
ge
neratio
n
wavelet tran
sform pre
p
ro
cessing suga
r partial
lea
s
t squares (P
LS) ,quantitative
analysi
s
mod
e
l
of the correlat
ion coefficie
n
t increa
se
d, the st
an
da
rd d
e
v
iation is
red
u
ce
d some
what, the seco
nd
gene
ration
wavelet tran
sfo
r
m of
spe
c
tra
l
data p
r
etrea
t
ment ca
n im
prove th
e p
r
e
d
iction
ability of
navel ora
nge
s su
gar PLS
model.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5414 – 54
19
5418
3.2. The Se
cond G
e
ner
a
tion
Wav
e
let Spec
tra
Data Pre
t
rea
t
ment E
ffe
cts on th
e Na
v
e
l
Orange Acidit
y
PLS Mod
e
l
Acco
rdi
ng to the method
mentioned i
n
se
ct
ion 1.2, collecte
d
the spe
c
tral
data of
predi
ction
n
a
vel orange
in collectio
n group
3, usin
g sta
n
dard
phy
sical and
ch
e
m
ical
measurement
method of section 1.
3 for determinin
g
corre
c
tion an
d grou
p 3 predictio
n sets
the
actual
a
c
idity of the
navel
orang
e, nav
el o
r
an
ge
aci
d
ity PLS qu
a
n
titative pre
d
i
c
tion
mod
e
l i
s
establi
s
h
ed, then u
s
e p
r
e
d
iction m
odel
to predi
ct
the acidity of the 20 n
a
vel oran
ge in thi
r
d
grou
p; used
the se
co
n
d
gen
eratio
n
wavele
t tra
n
sform for
navel orang
e sp
ectral
data
prep
ro
ce
ssin
g, build the navel ora
nge
acid
ity PLS quantitative predi
ction m
odel, then u
s
ing
model
s to pre
d
ict 3 grou
ps
that the acidity of
the
20 navel oran
ge. In Figure 4, Fi
gure 5, it is the
se
con
d
gene
ration wavele
t transform fo
r navel or
a
n
g
e
spe
c
tral da
ta preprocessing befo
r
e a
n
d
after the a
c
id
ity of discret
e
points di
strib
u
tion
of the p
e
rcentag
e of predi
cted val
ue and
actu
a
l
value:
Analysis th
e two
cha
r
ts
sh
ows that, the or
igin
al navel
oran
ge a
c
idit
y PLS model and the
se
con
d
ge
ne
ration
wavele
t correlatio
n
coeffici
ent of
the optimize
d
navel o
r
an
ge a
c
idity PLS
model, the st
anda
rd deviat
i
on of a calib
ration model
, t
he predictio
n
model of ro
ot mean squa
re
error value
s
a
s
sh
own in table:
By this table, it can been
see
n
that take
the se
con
d
generation
wavelet tran
sform of
navel orang
e
wa
s foun
de
d after the
spectra d
a
ta
pretreatment
acidity pa
rti
a
l lea
s
t sq
ua
res
(PLS) q
uantit
ative analysi
s
model
of
the correl
ation co
efficient
in
cre
a
se
d, the
sta
ndard d
e
viation
is lo
we
r a littl
e, With th
e seco
nd g
ene
ration
wavelet
tran
sform
of
sp
ectral d
a
ta prep
rocessi
ng
navel ora
nge
acidity PLS model imp
r
ove
d
accuracy a
nd relia
bility.
Figure 4. The
Acidity PLS
Model of Ori
g
inal
Navel Orang
e
Figure 5. The
Acidity PLS
Model of Nav
e
l
Oran
ge after
Secon
d
Gen
e
r
ation
Wavele
t
Pretreatm
ent
Table 3. The
Analysis
Re
sults of the Ori
g
i
nal an
d Second Ge
neration Wavel
e
t Optimization
Navel Orang
e Acidity Partial Lea
st Ssqu
are
s
(PLS) M
e
thod Mod
e
l
Number
of
samples
The corr
elation
coefficient model
Correc
t
i
on mod
e
l
standard deviation
Standard deviati
on
prediction model
PLS model of
original Brix
20 89.78
0.842
0.863
PLS model of
optimization Brix
20 92.18
0.691
0.738
4. Conclusio
n
In this study, use
d
a po
rta
b
le sp
ect
r
um
in
stru
ment, u
s
ed n
ear i
n
frared
diffuse reflection
spe
c
tru
m
det
ection te
chno
logy to acquired the nav
el oran
ge spe
c
tral data, elab
orated the ba
si
c
prin
ciple
of the second g
eneration
wa
velet tr
ansfo
rm, using th
e
se
con
d
gen
eration
wavel
e
t
transfo
rm of
spe
c
tral
data
prep
ro
ce
ssi
ng, com
par
ative analysi
s
it is con
c
lud
e
d that after the
se
con
d
g
ene
ration
wavelet
sp
ect
r
al d
a
ta prep
ro
ce
ssi
ng
sug
a
r
and
aci
d
ity of na
vel ora
nge
P
L
S
model predi
ctive ability is improve
d
. The opt
imized mo
de
l effectively eliminates the
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
The Na
vel O
r
ange Sug
a
r a
nd Acidity
Qu
antitative Pre
d
iction Mo
del
Optim
i
zation… (Zha
o Ke)
5419
instru
mentati
on, mea
s
u
r
e
m
ent conditio
n
, the influ
e
n
c
e of
sa
mple status
for sp
e
c
tral acqui
sition,
and by usi
ng
optimizatio
n model is
esta
blish
ed in
na
vel orang
e su
gar a
nd a
c
idit
y of partial least
squ
a
re
s a
nal
ysis mo
del.E
xperime
n
ts
show th
at the optimized
se
con
d
gen
e
r
ation
wavel
e
t
transfo
rm na
vel orang
e sugar a
nd aci
d
ity analysis
model ha
s g
ood predi
ctio
n ability, For navel
oran
ge suga
r and aci
d
ity accurately fore
ca
st prov
ide
s
a means of f
a
st non
de
stru
ctive detectio
n
.
Ackn
o
w
l
e
dg
ements
This wo
rk i
s
Suppo
rted
by the Prog
ram
of
Natu
ral Scie
nce F
ound
ation
of Jia
ngxi
ProvinceNo.20122BAB201026
Referen
ces
[1]
Mogh
imi A, Aghkh
ani M
H
. Vis-NIR Spectr
oscop
y
and
C
hemo Metric f
o
r the Pre
d
icti
on of So
lub
l
e
Soli
ds Conte
n
t and Aci
d
it
y
(P
H) of Ki
w
i
fr
uit.
Bio-system
Engineering.
20
1
0
; 106(3): 2
95-
302.
[2]
He
Y
ong, Li Xi
ao
Li. Near
Infrared
S
pect
r
oscop
y
w
i
th
ba
yb
err
y
var
i
e
t
ies.
Infrared
and
Mill
i
m
eter
Waves
. 2006;
25(3): 19
2-1
9
4
.
[3]
Z
hao K
e
,
Xio
n
g
Ya
n. T
he res
earch
of n
ear
i
n
frar
ed
spectro
scop
y
base
d
o
n
n
a
vel
ora
n
g
e
of rap
i
d
no
n-
destructive test
ing tech
no
log
y
.
Laser an
d infr
ared.
20
11; 41(
6): 649-6
52.
[4]
Liu Ya
nde, Su
n Xu
do
ng, Z
hang Ha
ili
an
g etal. Non
des
tr
uctive meas
urement of inter
nal q
ual
it
y
o
f
Nanfe
n
g
man
dari
n
fruit
b
y
char
ge c
o
u
p
l
ed
dev
ice
ne
ar i
n
frared
sp
ectroscop
y
.
N
ear Infrar
e
d
Spectrosco
py
. 201
0; 71(1): 10
-14.
[5]
Ke Z
hao. T
he
Rese
arch of G
r
a
y
Alg
o
rithm
and In
f
o
rmatio
n
Entrop
y in R
oute
Pl
an
nin
g
Optimizatio
n
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2012; 1
0
(8): 2027-
203
3
[6]
Che
ng-C
h
a
ng
Lie
n
, Chi
ng-
H
ua T
i
ng.
Sligh
t
free falli
ng
i
m
p
a
ct test for assessi
ng
gu
ava
matur
i
ty.
Procee
din
g
s of 2013 Spr
i
ng I
n
ternati
o
n
a
l C
onfere
n
ce
o
n
Agricult
ure an
d
F
ood Engi
nee
ring (AF
E
-S).
201
3.
[7]
Bobe
l
y
n E, Ser
ban AS, N
i
cu
M. Post harves
t
qual
it
y
of app
l
e
Pred
icted b
y
NI
R-spectrosc
o
p
y
: Stud
y o
f
the
effect of biol
og
ical
vari
abil
i
t
y
o
n
s
p
e
c
tra an
d Mo
d
e
l P
e
rformanc
e.
Post h
a
rve
s
t Biol
ogy
an
d
T
e
chno
logy.
2
010; 55(
3): 133
-143.
[8]
Liu Y
u
a
n
she
n
g
. Rese
arch
o
n
the K
e
y T
e
chni
que
of SP
C Exc
h
a
n
g
e
E
qui
pment B
a
s
ed o
n
F
P
GA.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2013; 1
1
(7): 3
809-
382
0.
Evaluation Warning : The document was created with Spire.PDF for Python.