Nea -in a ed spec oscopy and pa e n- ecogni ion p ocessing
o classi ying wines o wo I alian p o inces
A.G. Mignani a, L. Ciacche i a*, B. Go dillo b, A.A. Mencaglia a, M.L. González-Mi e b,
F.J. He edia b, A. Cichelli c
a CNR-Is i u o di Fisica Applica a “Nello Ca a a”
Via Madonna del Piano, 10 – 50019 Ses o Fio en ino (FI), I aly
b Lab. Colo y Calidad de Alimen os, Uni . de Se illa, Facul ad de Fa macia – 41012 Se illa, Spain
c Uni e si à degli S udi “G. D’Annunzio”, Dip. Economia – 65127 Pesca a, I aly
ABSTRACT
This pape p esen s an expe imen making use o he nea -in a ed spec um o dis inguishing he wines p oduced in
wo close p o inces o Ab uzzo egion o I aly. A collec ion o 32 wines was conside ed, 18 o which we e p oduced in
he p o ince o Chie i, while he o he 14 we e om he p o ince o Te amo. A con en ional dual-beam
spec opho ome e was used o abso p ion measu emen s in he 1300-1900 nm spec oscopic ange. P incipal
Componen Analysis was used o explo a i e analysis. Sco e maps in he PC1-PC2 o PC2-PC3 spaces we e ob ained,
which success ully g ouped he wine samples in wo dis inc clus e s, co esponding o Chie i and Te amo p o inces,
espec i ely. A modelling o dual-band spec oscopy was also p oposed, making use o wo LEDs o illumina ion and a
PIN de ec o ins ead o he spec ome e . These da a we e p ocessed using Linea Disc iminan Analysis which
demons a ed sa is ac o y classi ica ion esul s.
Keywo ds: wine, classi ica ion, spec oscopy, NIR, geog aphic o igin, mul i a ia e da a analysis
1. CLASSIFICATION OF WINES: WHY OPTICAL TECHNOLOGIES
The p omo ion o wines wi h a unique geog aphical conno a ion is conside ed a s a egic ac o o p o ec ing and
boos ing he Eu opean sha e o he wine ma ke . Labels bea ing denomina ion o o igin (PDO) and geog aphical
indica ion (PGI) adema ks a e o en used o highligh he peculia i ies o wines, o be e isibili y o consume s and
di e en ia ions wi h espec o simila p oduc s wi h lowe p ice. The e oi , as he speci ici y o place, has a
undamen al in luence on he wine quali y. The ole o he e oi includes no only he soil ype o ha egion, bu also
he clima e, he wea he , he ines and ineya ds, and any hing else ha can possibly di e en ia e one piece o land om
ano he , e en close.
In es iga ing and classi ying wine di e ences o au hen ica ion pu poses has been widely accomplished using
con en ional analy ical echniques such as high pe o mance liquid ch oma og aphy 1, gas ch oma og aphy 2, liquid
ch oma og aphy/mass spec ome y 3, and elemen al analysis 4.
Beside analy ical echniques sui able o labo a o y use only, op ical spec oscopy has eme ged as a apid and non-
des uc i e ool o quick measu emen s, since he en i e spec um om he ul a iole o he mid-in a ed is capable o
highligh ing e en minimal di e ences be ween wines 5, 6, 7, 8, 9, 10, 11, 12, 13. In ac , annins, phenolic compounds, o he
pigmen s, and he di e en con en o wa e , suga s, and e hanol, g ea ly in luence he op ical spec um and allow o
dis inguishing he di e en geog aphic a eas o wines. Spec oscopic da a a e usually p ocessed by means o mul i a ia e
da a analysis, demons a ing ha he combina ion o spec oscopy and chemome ics p o ides a mode n and
s aigh o wa d ool o wine classi ica ion and au hen ica ion. The nea -in a ed band, which is pa icula ly in o ma i e
o he concen a ion o wa e , suga s, and e hanol, demons a ed e ec i eness o assessing he chemical composi ion
and he a oma, and o moni o ing he e men a ion p ocess 14, 15, 16, 17, 18, 19.
This pape p esen s an expe imen which was ca ied ou in a small band o he nea -in a ed spec um o dis inguishing
he wines p oduced in wo close p o inces o Ab uzzo, which is a cen al egion o I aly. A con en ional dual-beam
* Email: l.ciacche i@i ac.cn .i – phone: +39 055 522 6322
Ad anced En i onmen al, Chemical, and Biological Sensing Technologies XI, edi ed by
Tuan Vo-Dinh, Robe A. Liebe man, Gün e G. Gaugli z, P oc. o SPIE Vol. 9106,
91060G · © 2014 SPIE · CCC code: 0277-786X/14/$18 · doi: 10.1117/12.2051914
P oc. o SPIE Vol. 9106 91060G-1
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx
spec opho ome e was used o abso p ion measu emen s in he nea -in a ed ange, showing he mos signi ican
di e ences among he a ious samples in he 1300-1900 nm band. P incipal Componen Analysis was i s ly used o
explo a i e analysis. Sco e maps in he PC1-PC2 o PC2-PC3 spaces success ully g ouped he wine samples in wo
dis inc clus e s, co esponding o Chie i and Te amo p o inces, espec i ely. Then, an inno a i e model was es ed,
which simula es he use o wo LEDs only o illumina ion, and a PIN de ec o ins ead o he spec ome e . Linea
Disc iminan Analysis was used o da a p ocessing, ob aining abou 12% classi ica ion e o .
2. THE WINE COLLECTION AND THE OPTICAL SPECTRA
Ab uzzo is a egion o cen al I aly. The Eas e n pa o e looks he Ad ia ic Sea, while he Wes e n pa includes he
G an Sasso moun ains. The wine p oduc ion o his egion is an impo an economic esou ce. Mos o he 4 million
hec oli es annually p oduced a e bea ing PDO and PGI labels: his means ha he p oduc ion is dis inc ly o ien ed
owa ds quali y a he han quan i y 20, 21. The expo ma ke is wo h o e 120 million eu o, a igu e ha showed a
g owing end and egis e ed a eco d o e he pas i e yea s. The mos popula ine ypes o his egion a e
Mon epulciano d’Ab uzzo, T ebbiano d’Ab uzzo, Peco ino and Cha donnay, and o he ines equen ly cul i a ed a e
Sangio ese, Me lo , Mal asia and Ce asuolo.
This s udy conside ed a selec ion o wines om Te amo and Chie i p o inces, espec i ely loca ed in he No -Eas and
Sou h-Eas pa o Ab uzzo, as shown in Figu e 1. Table I summa izes he cha ac e is ics o he collec ion, which
ep esen ed a signi ican egional igu e. I was made o 32 ed, osè and whi e wines o di e en ines, p oduced in
2007: 18 samples we e om he Chie i p o ince, while he o he 14 samples we e p oduced in he Te amo p o ince.
Code P o ince Village B and Va ie y Wine ype
1 CH O sogna O sogna Sangio ese whi e
2 CH O sogna O sogna Sangio ese whi e
3 CH O sogna O sogna Mon epulciano ed
4 CH O sogna O sogna Mon epulciano ed
5 CH O sogna O sogna Mal asia whi e
6 CH O sogna O sogna Cha donnay whi e
7 CH O sogna O sogna Peco ino whi e
8 CH O sogna O sogna Sangio ese ed
9 CH O sogna O sogna T ebbiano whi e
10 CH O sogna O sogna Mon epulciano ed
11 CH O ona O ona Cha donnay whi e
12 CH O ona O ona Peco ino whi e
13 CH O ona O ona T ebbiano whi e
14 CH O ona O ona Mon epulciano ed
15 CH O ona O ona Me lo ed
16 CH O ona O ona Ce asuolo osé
17 TE Sil i Sil i Mon epulciano ed
18 TE Sil i Sil i Cha donnay whi e
19 TE Sil i Sil i T ebbiano whi e
20 TE Sil i Sil i Peco ino whi e
21 TE Giuliano a Gio anPie o Ce asuolo osé
22 TE Giuliano a Gio anPie o Mon epulciano ed
23 TE Giuliano a Gio anPie o T ebbiano whi e
24 TE Giuliano a Gio anPie o Peco ino whi e
25 TE Canosa Nicola Ce asuolo osé
26 TE Cas ilen i SanLo enzo Ce asuolo osé
27 TE Cas ilen i SanLo enzo Peco ino whi e
28 TE Cas ilen i SanLo enzo Mon epulciano ed
29 TE Cas ilen i SanLo enzo T ebbiano whi e
30 TE Cas ilen i SanLo enzo Cha donnay whi e
31 CH Canosa Nicola Cha donnay whi e
32 CH To ino di Sang o Mucci Ce asuolo osé
Figu e 1. The loca ion o he Ab uzzo egion in I aly ( op), Table I. The wine collec ion.
and o he Te amo and Chie i p o inces (bo om).
P oc. o SPIE Vol. 9106 91060G-2
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx
1.4
1.2
Up
@ 0.8
Qo"op 0.6
0.4
0.2
L.
CH
- TE
1300 1400 1500 1600 1700 1800 1900
Wa eleng h (nm )
0.25
0.2
0.15
0.1
m
0.05
oJo
-0.05
-0.1
- PC 1
-PC2PC 3
0.1100 1400 1500 1600 1700
Wa eleng h (nm ) 1800 1900
6
5
4
2
1
O
14
12
1
0.8
SD
_§ 00.6
0 4
02
-RedRosé
500 1000 1500 2000
-Red
- -RoséWhi e
g00 1000 1200 1400 1600 1800 2000
A con en ional dual-beam spec opho ome e was used o abso p ion spec oscopy measu emen s. Qua z cu e es wi h
1 mm ligh pa h we e used, wi h wa e in he e e ence channel. Figu e 2- op shows he abso p ion spec a in he en i e
isible and nea -in a ed ange. The isible spec a a e clea ly domina ed by he colo o wines, and he e o e we e
dis ega ded. The spec oscopic di e ences in he nea -in a ed band a e highligh ed in Figu e 2-bo om: he mos
signi ican di e ences among he a ious samples a e in he 1300-1900 nm ange. These di e ences a e ela ed o he
i s o e one o he OH s e ch o wa e , and a combina ion o s e ch and de o ma ion o he OH g oup in wa e and
e hanol 22, 23.
Figu e 2. Wide band abso p ion spec a o all wines ( op),
and nea -in a ed band only (bo om).
Figu e 3. Selec ed nea -in a ed band o da a
p ocessing ( op), and ela ed PCA loadings (bo om).
3. GEOGRAPHIC CLASSIFICATION
The P incipal Componen Analysis (PCA), which is one o he mos popula echniques o explo a i e analysis and da a
dimensionali y educ ion, was used o p ocessing he spec oscopic da a in he 1300-1900 nm ange. PCA p o ides new
a iables and coo dina es o iden i ying he wine samples in a 2D o 3D map. The coe icien s gi ing he weigh o each
a iable in he new space a e called loadings. The new a iables a e called P incipal Componen s (PCs), and ha e he
ollowing p ope ies:
- PCs a e mu ually unco ela ed (o hogonali y).
- 1s PC (PC1) has he la ges a iance among all possible linea combina ion o he s a ing a iables.
- PCn has he la ges a iance among all linea combina ion o he s a ing a iables ha a e o hogonal o PC1 ...
PC(n - 1).
P oc. o SPIE Vol. 9106 91060G-3
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx
e
N
10
5
0
N-5
Ua
-10
iliik
_119 q/43 _
98 4412
q73q415 q8
°'qaq192
qe
_
4
(4,0/ _
oCH
0 TE
010 20
PC1 (61.3%) 30 40
10
5
0
-10
15-1
4°'
01
09
°d <3z
q
D2
147
97
_
95
9
-`°4e
118
42
9.0
92 98
9i C i 95
q4
o
0CH
TE
-10 -5 0
PC2 (26.8%) 510
This means ha high o de PCs has li le weigh in dis inguishing he samples, and can be dis ega ded wi h li le loss o
in o ma ion. The loading plo s a e use ul o in e p e he sco e map: hey show wha a iables a e impo an o a gi en
PC: a iable wi h 0 loading has no impo ance, a a iable wi h high (posi i e o nega i e) loading is impo an o
di e en ia ing he wines 24, 25.
Figu e 3-bo om shows he loading o PCA p ocessing o wine spec a in he1300-1900 nm band:
- PC1 has a nea ly cons an beha io , and is ela ed o he luc ua ions o he baseline.
- PC2 exp esses he di e ence be ween he abso bance a 1400 nm and 1500 nm. No e ha hese wo
wa eleng hs a e posi ioned espec i ely on he ascending and descending slope o he peak a 1450 nm.
Usually, his beha io is ela ed o a shi in he cen al wa eleng h o he peak.
- PC3 has he maximum loading in he cen e o he peak, and is clea ly ela ed o he peak heigh .
Figu e 4 shows he sco e maps in he PC1-PC2 and PC2-PC3 spaces which success ully g ouped he wine samples in
wo dis inc clus e s, co esponding o he Chie i and Te amo p o inces, espec i ely.
Figu e 4. Resul s o PCA p ocessing o spec oscopic da a in he 1300-1900 nm band: clus e ing acco ding o he Chie i and Te amo p o inces.
4. PREDICTIVE MODEL FOR DUAL-BAND SPECTROSCOPY
Since he disc imina ing componen PC2 is ela ed o he di e en ial abso bance be ween 1400 and 1500 nm, an a emp
was made o classi ying he wines om he wo p o ince simply using wo LED o illumina ion, and a PIN de ec o
ins ead o he spec ome e . This dual-band scheme could be in e es ing o implemen ing a low-cos de ice o
consume s. Indeed, he g owing in e es o consume s o use cheap de ices o sel -assessmen o ood quali y, was he
mo i a ion ha inspi ed his wo k.
Fo his scope, he spec a o Figu e 3- op we e con olu ed wi h Gaussian weigh ing unc ions, hus modelling he
emission o comme cially-a ailable LEDs. Two bands we e chosen, i ing he bands o he mos e iden spec oscopic
di e ences. Then, by calcula ing he in eg al o he abso p ion spec a in hese bands, each wine sample was iden i ied
by wo coo dina es. The cha ac e is ics o comme cially a ailable LEDs we e conside ed in he simula ion 26. Because
LEDs emi ing exac ly a 1400 nm and 1500 nm we e no eadily a ailable, we made a comp omise by choosing LEDs
cen e ed a 1450 nm, and 1480 nm, espec i ely. Since he emission bands o hese LEDs we e wide han 100 nm, he
use o no ch il e s was conside ed, so as o ob ain FWHM=12 nm o bo h LEDs, as displayed in Figu e 5-le .
The da a se was p ocessed using he Linea Disc iminan Analysis (LDA), which is a obus and eliable echnique o
au oma ic objec classi ica ion 27. Like PCA, LDA p ojec s a high-dimensional pa e n on o a subspace o smalle
dimension, bu he axes o p ojec ion a e chosen using a di e en c i e ion. Ac ually, LDA is a ool ha is speci ically
sui ed o iden i ica ion, and looks o hose a iables ha show a la ge sp ead among di e en clus e s (in e -class
a iance), bu limi ed a iance wi hin each clus e (in a-class a iance). Gi en an N-class p oblem, he LDA ex ac s
om he da a ma ix N-1 Disc imina ing Func ions (DFs), which co espond o P incipal Componen s in he PCA, bu
show a be e esolu ion wi h ega d o poo ly-sepa a ed clus e s. In ou case, being a wo-class p oblem, we ex ac ed
one DF only.
P oc. o SPIE Vol. 9106 91060G-4
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx
-o *---¢- *mew -cam--
1450 1500
Wa eleng h (nm ) 1550
0.5
0
00504
decision bo de
-iii -mom --
RCHlE
0.045 0.05
DF 0.055
Figu e 5- igh shows he esul s o he LDA p ocessing: he co ec classi ica ion a e using all samples o calib a ion
was 81.25%. The co ec ly classi ied wines we e 15 om Te amo ou o 18, and 11 om Chie i ou o 14. The lea e-
one-ou c oss- alida ion p ocedu e ga e a success a e o 78.1%, qui e close o he calib a ion alue.
Figu e 5. The model o dual-band spec oscopy – Wine abso p ion spec a in he limi ed 1400-1500 nm band, including he emission spec a
o wo LEDs (le ). Resul s o LDA p ocessing ( igh ).
5. PERSPECTIVES
Abso p ion spec oscopy in he nea -in a ed band, combined o mul i a ia e da a p ocessing, demons a ed e ec i eness
o wine au hen ica ion. Wines p oduced in he Ab uzzo egion o cen al I aly we e success ully classi ied acco ding o
wo p o inces, Te amo and Chie i, espec i ely. Mo eo e , a model o dual-band spec oscopy was p oposed, making
use o wo LEDs o illumina ion and a PIN de ec o ins ead o he spec ome e . Also his model showed a good
classi ica ion acco ding o p o inces, demons a ing he po en ials o implemen ing a low-cos de ice o consume use.
ACKNOWLEDGEMENTS
The Conseje ía de Inno ación, Ciencia y Emp esa (Jun a de Andalucía) is acknowledged o pa ial inancial suppo
(IAC07-I-1664).
REFERENCES
1 S.A. Belloma ino, X.A. Conlan, R.M. Pa ke , N.W. Ba ne , M.J. Adams, “Geog aphical classi ica ion o some
Aus alian wines by disc iminan analysis using HPLC wi h UV and Chemiluminescence de ec ion”, Talan a 80, 833-
838 (2009).
2 S.M. Rocha, E. Coelho, J. Z oslìko à, I. Delgadillo, M.A. Coimb a, “Comp ehensi e wo-dimensional gas
ch oma og aphy wi h ime-o - ligh mass spec ome y o mono e penoids as a powe ul ool o g ape o igin
aceabili y”, Jou nal o Ch oma og aphy A, 1161, 292-299 (2007).
3 L. Jai z, K. Siegl, R. Ede , G. Rak, L. Ab anco, G. Koellenspe ge , S. Hann, “LC-MS/MS analysis o phenols o
classi ica ion o ed wine acco ding o geog aphic o igin, g ape a ie y and in age”, Food Chemis y 122, 366-372
(2010).
4 J.M. Ju ado, A. Alcàza , A. Palacios-Mo illo, F. De Pablos, “Classi ica ion o Spanish DO whi e wines acco ding o
hei elemen al p o ile by means o suppo ec o machines”, Food Chemis y, 135, 898-903 (2012).
5 D. Cozzolino, W.U. Cynka , P.A. Smi h, “Can spec oscopy geog aphically classi y Sau ignon Blanc wines om
Aus alia and New Zealand?”, Food Chemis y, 126, 673-678 (2011).
6 D. Cozzolino, M. Halds ock, R.G. Dambe gs, W.U. Cynka , P.A. Smi h, “Mid in a ed spec oscopy and mul i a ia e
analysis: a ool o disc imina e be ween o ganic and non-o ganic wines g own in Aus alia”, Food Chemis y, 116, 761-
765 (2009).
7 L. Liu, D. Cozzolino, W.U. Cynka , R.G. Dambe gs, L. Janik, B.K. O’Neill, C.B. Colby, M. Gishen, “P elimina y
s udy on he applica ion o isible-nea in a ed spec oscopy and chemome ics o classi y Riesling wines om di e en
coun ies”, Food Chemis y, 106, 781-786 (2008).
P oc. o SPIE Vol. 9106 91060G-5
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx

8 C.J. Be in, R.G. Dambe gs, A.J. Fe gusson, D. Cozzolino, “Va ie al disc imina ion o Aus alian wines by means o
mid-in a ed spec oscopy and mul i a ia e analysis”, Analy ica Chimica Ac a, 621, 19-23 (2008).
9 M. U bano, M.D. Luque de Cas o, P-M. Pé ez, J. Ga cìa-Olmo, M.A. Gòmez-Nie o, “Ul a iole - isible spec oscopy
and pa e n ecogni ion me hods o di e en ia ion and classi ica ion o wines”, Food Chemis y, 97, 166-175 (2006).
10 L. Liu, D. Cozzolino, W.U. Cynka , M. Gishen, C.B. Colby, “Geog aphic classi ica ion o Spanish and Aus alian
Tempo illo ed wines by isible and nea -in a ed spec oscopy combined wi h mul i a ia e analysis”, Jou nal o
Ag icul u al and Food Chemis y, 54, 6754-6759 (2006).
11 C.J. Be in, A.J. Fe gusson, W.B. Pe ey, L.J. Janik, D. Cozzolino, “De elopmen o a apid inge p in ing sys em o
wine au hen ici y by mid-in a ed spec oscopy”, Jou nal o Ag icul u al and Food Chemis y, 54, 9713-9718 (2006).
12 D. Cozzolino, H.E. Smy h, K.A. La ey, W. Cynka , L. Janik, R.G. Dambe gs, I.L. F ancis, M. Gishen, “Rela ionship
be ween senso y analysis and nea in a ed spec oscopy in Aus alian Riesling and Cha donnay wines”, Analy ica
Chimica Ac a, 539, 341-348 (2005).
13 D. Cozzolino, H.E. Smy h, M. Gishen, “Feasibili y s udy on he use o isible and nea -in a ed spec oscopy oge he
wi h chemome ics o disc imina e be ween comme cial whi e wines o di e en a ie al o igins”, Jou nal o
Ag icul u al and Food Chemis y, 51, 7703-7708 (2003).
14 W. Cynka , D. Cozzolino, R.G. Dambe gs, “The e ec o sample s o age and homogenisa ion echniques on he
chemical composi ion and nea in a ed spec a o whi e g apes”, Food Resea ch In e na ional, 42, 653-658 (2009).
15 H.E. Smy h, D. Cozzolino, W.U. Cynka , R.G. Dambe gs, M. Se on, M. Gishen, “Nea in a ed spec oscopy as a
apid ool o measu e ola ile a oma compounds in Riesling wine: possibili ies and limi s”, Analy ical Bioanaly ical
Chemis y, 390, 1911-1916 (2008).
16 D. Cozzolino, W.U. Cynka , R.G. Dambe gs, M.D. Me cu io, P.A. Smi h, “Measu emen o condensed annins and
d y ma e in ed g ape homogena es using nea in a ed spec oscopy and pa ial leas squa es”, Jou nal o Ag icul u al
and Food Chemis y, 56, 7631-7636 (2008).
17 D. Cozzolino, M.J. Kwia owski, R.G. Dambe gs, W.U. Cynka , L.J. Janik, G. Skou oumounis, M. Gishen, “Analysis
o elemen s in wine using nea in a ed spec oscopy and pa ial leas squa e eg ession”, Talan a, 74, 711-716 (2008).
18 D. Cozzolino, R.G. Dambe gs, L. Janik, W.U. Cynka , M. Gishen, “Analysis o g apes and wine by nea in a ed
spec oscopy – a e iew”, Jou nal o Nea In a ed Spec oscopy, 14, 279-289 (2006).
19 J. Fe nández-No ales, M.I. López, M.T. Sánchez. J.A. Ga cía, J. Mo ales, “A easibili y s udy on he use o a
minia u e ibe op ic NIR spec ome e o he p edic ion o olumic mass and educing suga s in whi e wine
e men a ions”, Jou nal o Food Enginee ing, 89, 325-329 (2008).
20 Web: h p://www.conso zio- iniab uzzo.i /
21 Web: h p://www.i aly-wine- ou s.ne /en/wine ou ab uzzo/wine ou ab uzzo/wine ou ab uzzo.php
22 D. Cozzolino, M. Pa ke , R.G. Dambe gs, M. He de ich, M. Gishen, “Chemome ics and isible-nea in a ed
spec oscopic moni o ing o ed wine e men a ion in a pilo scale”, Bio echnology and Bioenginee ing, 95, 1101-1107
(2006).
23 D. Cozzolino, M.J. Kwia kowski, M. Pa ke , W.U. Cynka , R.G. Dambe gs, M. Gishen, M.J. He de ich, “P edic ion o
phenolic compounds in ed wine e men a ions by nea in a ed spec oscopy”, Analy ica Chimica Ac a, 513, 73-80
(2004).
24 M.J. Adams, [Chemome ics in Analy ical Spec oscopy], The Royal Socie y o Chemis y, Camb idge, UK (1995).
25 J.E. Jackson, [A Use ’s Guide o P incipal Componen s], J. Wiley & Sons Inc., Hoboken, NJ (2003).
26 Web: h p://www. oi hne -lase .com/led_di e se.h ml
27 B.G.M. Vandegins e, D.L. Massa , L.C.M. Buydens, S. De Jong, D.J. Lewi, J. Smeye s-Ve beke, [Handbook o
Chemome ics and Qualime ics], Else ie Science BV, Ams e dam – Chap e 33 (1998).
P oc. o SPIE Vol. 9106 91060G-6
Downloaded F om: h p://p oceedings.spiedigi allib a y.o g/ on 05/11/2016 Te ms o Use: h p://spiedigi allib a y.o g/ss/Te msO Use.aspx