Beruflich Dokumente
Kultur Dokumente
Brix
and Pol are usually measured in the laboratory using juice
samples. Unfortunately, replicating this laboratory method in
the eld is very difcult due to the absence of reliable sensor
and the difculties obtaining sufcient juice samples. There-
fore, an ability to predict sugarcane quality by directly scan-
ning stalk skin using spectroscopic methods would be an
attractive alternative. Spectroscopic methods have widely
been applied to non-destructively predict internal crop quality
based on skin scanning of a number of crops, e.g. kiwifruit
(Moghimi, Aghkhani, Sazgarnia, & Sarmad, 2010), apples (Lu,
2004; Park, Abbott, Lee, Choi, & Choi, 2003) and pineapple
(Chia, Rahim, & Rahim, 2012; Guthrie & Walsh, 1997).
For eld application, however, a portable, robust and low
cost spectrometer is preferable (Montes et al., 2006). Lately,
the development of a portable visible and shortwave near
infrared (Vis/SWNIR) instrument with a wavelength range
from 350 to 1100 nmappears promising because the bands are
ascribed to the third and fourth overtones of OeH and CeH
stretching modes and the instrument is low-cost and portable
enough for in-eld measurements (Walsh, Guthrie, & Burney,
2000). The use of this spectrometer, coupled with the use of
multivariate statistical techniques could create the possibility
to intensively and precisely map crop quality over large parcel
of cropping land (Menesatti et al., 2010). Several studies have
been successfully conducted in the eld using a spectrometer
mounted on a harvester for different crops, including maize
(Montes et al., 2006), forage (Digman & Shinners, 2008) and
sugar beet (Panigrahi & Hofman, 2003).
Unfortunately, there are no reported studies regarding the
skin scanning method for quality prediction in sugarcane
industry. This method works based on the principle that when
a light beam hits the stalk, a small fraction is reected at the
surfaceas a specular reectanceandthe rest will penetrateinto
fruit tissues. In the tissues, photons are absorbed or migrate in
different directions where radiationwill be scattered backward
to the surface as diffuse reectance, while the remaining ra-
diation migrates forward into the tissues as absorbance (Qing,
Ji, & Zude, 2007). Light absorption is related to certain chemi-
cal constituents, such as sugar, acid, water, etc. (Williams &
Norris, 2001). A reectance spectrometer measures the aggre-
gate amount of light reected from a sample, from which light
absorption may be estimated and then related to certain
chemical constituents of the crop (Lu, 2004).
For mapping purposes, the crop quality data must also be
classied into suitable quality classes. The articial neural
network (ANN) is a well-known non-linear method which
could provide a robust classication model (Lee et al., 2010).
ANN is a multivariate statistical data analysis technique which
often gives higher recognition and prediction probability than
statistical classication algorithms. Inspired by biological ner-
vous system, the neural networks are composed of a number of
elements operating in parallel. By adjusting the weights (con-
nections) between elements, a neural network can perform
several functions, such as prediction, data ltration, data
conceptualization, classication, and data association (Haykin,
1994). The applications of ANN to classify crop quality have
been reported by Wu, Feng, Zhang, and He (2008) and Xing,
Symons, Shahin, and Hatcher (2010). In this study, ANN
models were applied to classify sugarcane quality based on
several
Brix values. Thus, the specic objectives of this study
were (1) to investigate the feasibility of using Vis/SWNIR to
predict the quality level based on skin scanning; (2) to classify
sugarcane quality based on ve
Brix classes using ANN.
2. Material and methods
2.1. Sample preparation
A total of 291 internodes were extracted from 22 sugarcane
stalk samples. The stalk samples were collected from the
research station of Bureau of Sugar Experimental Station
Nomenclature
A Absorbance spectra
ANN Articial neural network
ASCII American Standard Code for Information
Interchange
ASD Analytical spectral device
BSES Bureau of Sugar Experimental Station
CCS Commercial cane sugar
FOV Field-of-view
FWHM Full width at half maximum
H High
L Low
LVs Latent variables
M Medium
MSC Multiplicative scatter correction
PA Precision agriculture
PCA Principal component analysis
PCs Principal components
PLS Partial least square
R Reectance spectra
R
2
coefcient of determination
RMSEP Root mean square error of prediction
SEC Standard error of calibration
SEP Standard error of prediction
SNV Standard normal variate
SSC Soluble solids content
Vis/SWNIR Visible/shortwave near infrared
VisNIR Visible near infrared
b i os y s t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 155
(BSES), Bundaberg, Queensland, inMay 2012. The stalk samples
were of a commercial variety cut froma propagationblock used
to source materials for plant breeding trials. They were a plant
crop that was planted in September 2011 (8 months old). The
crop was grown under commercial conditions with the fertil-
isation based on soil test and the six easy steps nutrition
guidelines. The stalks belong to commercial variety trials rep-
resenting three different maturity stages, namely early-
maturing (Q155), mid-maturing (Q208) and late-maturing
(Q190) crops. The selection of these three varieties was to
ensure that the models developed in this study cover a wide
range of
Brix (7.6e22.2) which is representative during har-
vesting. Whole stalk samples were rst topped and cut into
individual internode using a cutter. Eachinternode sample was
scanned at four different scanning points from bottom to the
top (Fig. 1). Each section was imaginarily labelled as S1, S2, S3
andS4 following the sequence fromS1 to S4 (bottomto the top).
2.2. Reectance measurement
The reectance measurement method was chosen for this
study because this method is the easiest to perform as it re-
quires no contact with the fruit and light levels are relatively
high (Schaare & Fraser, 2000). Reectance measurement
methods have been successfully used to measure crops
quality by Chia et al. (2012) and Digmanand Shinners (2008). In
this study, a handheld Vis/NIR spectroradiometer (FieldSpec
HandHeld and FieldSpec Pro FR, 325e1075 nm, Analytical
Spectral Devices (ASD), Inc., Boulder, CO, USA) was used to
collect spectra data fromthe skin surface over the wavelength
range of 325e1075 nm in 1.5 nm intervals. The equipment
used has a spectral resolution (full width at half maximum
(FWHM) of a single emission line) of approximately 3 nm at
around 700 nm. The scanning was undertaken using the 25
Brix
values) and the variable matrix X (spectral data). Thus, PLS
method was used in this study to interpret the spectra and
develop both calibration and prediction models for sugarcane
Brix
values was obtained due to the inclusion of three different
varieties with different maturity stages. The range of
Brix
values for internode samples fromthe top to the bottomof the
Q155, Q208 and Q190 varieties were 7.6e22.2, 8e21.4 and 8e21,
respectively. The variation of
Brix values along the stalk for
each variety as measured from each internode sample is
shown in Fig. 2. The downward trend of
Brix values along the
stalk is consistent for all varieties. The graph also showed that
different varieties have different internode numbers. It can be
seen that the early maturing variety (Q155) had higher
Brix
values than other varieties especially for the several in-
ternodes at the bottom of the stalks.
The example of raw reectance spectra (R) and absorbance
spectra (A) of three samples having high (22
Brix), medium
Table 1 e Classication table for sugarcane
Brix.
Class
Brix range
High (H) 19.3e22.2
Medium high (MH) 16.5e19.2
Medium (M) 13.5e16.4
Medium low (ML) 10.5e13.4
Low (L) 7.6e10.4
b i os y s t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 157
(18
Brix) and low (14.2
Brix)
Brix values as measured by Vis/
SWNIR are shown in Fig. 3(a) and (c) respectively. In each
gure, no obvious difference could be seen in the shape of the
spectra for different
Brix values for both gures. However,
a baseline shift problem existing in raw reectance and
absorbance spectra has been eliminated using MSC method as
shown in Fig. 3(b) and (d) respectively.
3.2. The prediction of sugarcane
Brix
PLS models were developed using the pretreated reectance
spectra by MSC technique. The performance of the nal PLS
calibration models could be evaluated by the standard error of
calibration (SEC), the coefcient of determination for calibra-
tion (R
2
), the standard error of prediction (SEP), and the coef-
cient of determination for prediction (R
2
). A proper model
should have a low SEC, SEP and root mean square error of
prediction (RMSEP) and a high coefcient of determination for
both prediction and calibration models. The values of these
indices for PLS models developed using reectance and
absorbance spectra are shown in scatter plots in Fig. 4(a) and
(b) respectively. In both gures, the ordinate and abscissa
represent the predicted and measured values of the
Brix. The
R
2
and RMSEP values for reectance spectra were 0.91 and 1.41
while for absorbance spectra were 0.89 and 1.51 respectively.
The prediction performance of PLS models for both reec-
tance and absorbance spectra showed good agreement be-
tween the reference and estimated values. Shenk and
Westerhaus (1996) suggested that an R
2
value greater than
0.9 indicates excellent quantitative information of the models.
From Fig. 4, it can be seen that the PLS model developed
with reectance spectra performed better than the model
developed with absorbance spectra as indicated by R
2
and
RMSEP. This is reasonable nding since the absorbance spectra
is related to the presence of chemical components such as
sugars, while the reectance spectra of fruit contain informa-
tion on both the absorption as well as scattering properties of
the tissue (Nicola et al., 2008). Since this study was performed
on skin surface, scattering properties of the tissue would also
have contributed to the model performance. The ten LVs used
in PLS model were also found to be sufcient for predicting
B
r
i
x
Internode Number
Q155 Q208 Q190
Fig. 2 e Typical average
Brix values of internodes for
different sugarcane varieties.
Table 2 e Statistical characteristics of
Brix in sugarcane
stalks for calibration and validation.
Model Sample no Min Max Mean SD
Calibration 200 7.5 22.2 17.86 3.04
Prediction 71 8.2 22 17.83 2.93
SD Standard deviation.
b i o s ys t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 158
10
20
30
40
50
400 500 600 700 800 900 1000
R
e
f
l
e
c
t
a
n
c
e
(
%
)
Wavelength (nm)
22 Brix 18 Brix 14.2 Brix
(a) (b)
(d)
15
20
25
30
35
40
45
400 500 600 700 800 900 1000
R
e
f
l
e
c
t
a
n
c
e
(
%
)
Wavelength (nm)
22 Brix 18 Brix 14.2 Brix
2
3
4
5
6
7
8
400 500 600 700 800 900 1000
A
b
s
o
r
b
a
n
c
e
L
o
g
(
1
/
R
)
Wavelength (nm)
22 Brix 18 Brix 14.2 Brix (c)
2
3
4
5
6
7
400 500 600 700 800 900 1000
A
b
s
o
r
b
a
n
c
e
L
o
g
(
1
/
R
)
Wavelength (nm)
22 Brix 18 Brix 14.2 Brix
Fig. 3 e Typical sugarcane spectrum at different
Brix values for (a) raw reectance spectra; (b) raw reectance spectra
pretreated with MSC; (c) raw absorbance spectra; (d) absorbance data pretreated with MSC.
12
14
16
18
20
22
12 14 16 18 20 22
P
r
e
d
i
c
t
e
d
(
B
r
i
x
)
Reference (Brix)
R
RMSEP
n = 71
slope = 0.66
2
= 0.89
= 1.51
(a)
10
12
14
16
18
20
22
10 12 14 16 18 20 22
P
r
e
d
i
c
t
e
d
(
B
r
i
x
)
Reference (Brix)
(b)
R
2
n = 71
Slope = 0.72
= 0.91
RMSEP = 1.41
Fig. 4 e Prediction versus reference values for prediction models of sugarcane
Brix using (a) absorbance spectra and (b)
reectance spectra.
b i os y s t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 159
4. Conclusions
Vis/SWNIR is a portable, lowcost and non-destructive method
which was applied in this study to predict sugarcane
Brix
based on stalk scanning. The PLS models showed a good
agreement between the reference and estimated values for
both reectance and absorbance spectra with R
2
were 0.91 and
0.89 respectively. The ANN used to classify
Brix into several
quality classes had yielded good classication performance
ranging from 50 to 100% accuracy with overall accuracy of
83.1%. Overall, this study has demonstrated that the Vis/
SWNIR spectroscopy technique coupled with ANN has the
potential to be used for on-line quality measurement to full
the requirement of PA. This study also demonstrated that
quality prediction based skin scanning using a Vis/SWNIR
spectroscopy is feasible. This technique will be a better
alternative to current quality measurement which is based on
juice samples. Thus, the next study will focus on how to apply
this technique to map sugarcane
Brix during harvesting in
the eld. However, an effort is required to design a proper
sampling mechanism and installation conguration before
this method can be applied on a harvester. Spectral variations
due to different varieties, planting regions, seasons, weather
and cultural conditions also need to be considered in the
future works.
Acknowledgements
The authors acknowledge the nancial supports provided by
Ministry of Higher Education, Malaysia and National Center
for Engineering in Agriculture (NCEA), Toowoomba, Australia.
The authors also thank BSES Limited, Bundaberg, for
providing samples and equipment.
r e f e r e n c e s
Arana, I., Jare n, C., & Arazuri, S. (2005). Maturity, variety and
origin determination in white grapes (Vitis vinifera L.) using
NIRS. Journal of Near Infrared Spectroscopy, 13, 349e357.
ASD. (2005). Handheld spectroradiometer: Users guide version 4.05.
Suite A Boulder, USA: Analytical Spectral Devices, Inc.
Bramley, R. G. V. (2009). Lessons from nearly 20 years of precision
agriculture research, development, and adoption as a guide to
its appropriate application. Crop & Pasture Science, 60, 197e217.
Bramley, R. G. V., Panitz, J. H., Jensen, T., & Baillie, C. (2012).
Within block spatial variation in CCS e another potentially
important consideration in the application of precision
agriculture to sugarcane production. Proceedings of the
Australian Society of Sugar Cane Technologists, 34, 1e8.
Chia, K. S., Rahim, H. A., & Rahim, R. A. (2012). Prediction of
soluble solids content of pineapple via non-invasive low cost
visible and shortwave near infrared spectroscopy and articial
neural network. Biosystems Engineering, 113, 158e165.
Cox, G., Harris, H., Pax, R., & Dick, R. (1996). Monitoring cane yield
by measuring mass ow rate through the harvester.
Proceedings of the Australian Society of Sugar Cane Technologists,
152e157.
Digman, M. F., & Shinners, K. J. (2008). Real time moisture
measurement on a forage harvester using near infrared
reectance spectroscopy. Transactions of the ASABE, 51,
1801e1810.
Guthrie, J., & Walsh, K. (1997). Non-invasive assessment of
pineapple and mango fruit quality using near infra-red
spectroscopy. Australian Journal of Experimental Agriculture, 37,
253e263.
Haykin, S. (1994). Neural network: A comprehensive foundation. New
York, NY: Macmillan.
Johnson, R. M., & Richard, E. P., Jr. (2005). Sugarcane yield,
sugarcane quality, and soil variability in Louisiana. Agronomy
Journal, 97, 760e771.
Khuriyati, N., & Matsuoka, T. (2004). Near infrared transmittance
method for nondestructive determination of soluble solid
content in growing tomato fruits. Environmental Control in
Biology, 42, 217e223.
Lee, W. S., Alchanatis, V., Yang, C., Hirafuji, M., Moshou, D., &
Li, C. (2010). Sensing technologies for precision specialty crop
production. Computers and Electronics in Agriculture, 74, 2e33.
Liu, F., He, Y., & Wang, L. (2008). Comparison of calibrations for
the determination of soluble solids content and pH of rice
vinegars using visible and short-wave near infrared
spectroscopy. Analytica Chimica Acta, 610, 196e204.
Lu, R. (2004). Multispectral imaging for predicting rmness and
soluble solids content of apple fruit. Postharvest Biology and
Technology, 31, 147e157.
Magalha es, P. S. G., & Cerri, D. G. P. (2007). Yield monitoring of
sugar cane. Biosystems Engineering, 96(1), 1e6.
Menesatti, P., Antonucci, F., Pallottino, F., Roccuzzo, G.,
Allegra, M., Stagno, F., et al. (2010). Estimation of plant
nutritional status by Vis/NIR spectrophotometric analysis on
orange leaves [Citrus sinensis (L) Osbeck cv Tarocco]. Biosystems
Engineering, 105, 448e454.
Moghimi, A., Aghkhani, M. H., Sazgarnia, A., & Sarmad, M. (2010).
Vis/NIR spectroscopy and chemometrics for the prediction of
soluble solids content and acidity pH of kiwifruit. Biosystems
Engineering, 106(3), 295e302.
Mohan, L. A., Karunakaran, C., Jayas, D. S., & White, N. D. G.
(2005). Classication of bulk cereals using visible and NIR
reectance characteristics. Canadian Biosystems Engineering,
47(7), 7e14.
Montes, J. M., Utz, H. F., Schipprack, W., Kusterer, B.,
Muminovic, J., Paul, C., et al. (2006). Near-infrared
spectroscopy on combine harvesters to measure maize grain
dry matter content and quality parameters. Plant Breeding, 125,
591e595.
Nicola, B. M., Verlinden, B. E., Desmet, M., Saevels, S., Saeys, W.,
Theron, K., et al. (2008). Time-resolved and continuous wave
NIR reectance spectroscopy to predict soluble solids content
and rmness of pear. Postharvest Biology and Technology, 47,
68e74.
Ns, T., Isaksson, T., Fearn, T., & Davies, T. (2004). A user-friendly
guide to multivariate calibration and classication. Charlton,
Chichester, UK: NIR Publications.
Table 3 e ANN classication results.
Class
Brix
range
No. of test
sample in
each group
No. of correct
classication
Accuracy
(%)
High (H) 19.3e22.2 30 30 100
Medium
High (MH)
16.5e19.2 23 14 60.9
Medium (M) 13.5e16.4 11 10 90.9
Medium
Low (ML)
10.5e13.4 4 2 50
Low (L) 7.6e10.4 3 3 100
Total 71 59 83.1
b i o s ys t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 160
Panigrahi, S., & Hofman, V. (2003). On the go sugar sensor for
determining sugar content during harvesting. U.S. Patent No.
6,624,888.
Park, B., Abbott, J. A., Lee, K. J., Choi, C. H., & Choi, K. H. (2003).
Near-infrared diffuse reectance for quantitative and
qualitative measurement of soluble solids and rmness of
delicious and gala apples. Transactions of the ASAE, 46(6),
1721e1731.
Peirs, A., Lammertyn, J., Ooms, K., & Nicola, B. M. (2001).
Prediction of the optimal picking date of different apple
cultivars by means of VIS: NIR-spectroscopy. Postharvest
Biology and Technology, 21, 189e199.
Price, R. R., Johnson, R. M., Viator, R. P., Larsen, J., & Peters, A.
(2011). Fiber optic yield monitor for a sugarcane harvester.
Transactions of the ASABE, 54(1), 31e39.
Purcell, D. E., Leonard, G. J., OShea, M. G., & Kokot, S. (2005).
A chemometrics investigation of sugarcane plant properties
based on the molecular composition of epicuticular wax.
Chemometrics and Intelligent Laboratory Systems, 76, 135e147.
Qing, Z., Ji, B., & Zude, M. (2007). Predicting soluble solid content
and rmness in apple fruit by means of laser light
backscattering image analysis. Journal of Food Engineering, 82,
58e67.
Ruan, R., Almaer, S., & Zhang, J. (1995). Prediction of dough
rheological properties using neural networks. Cereal Chemistry,
72(3), 308e311.
Schaare, P. N., & Fraser, D. G. (2000). Comparison of reectance,
interactance and transmission modes of visible-near infrared
spectroscopy for measuring internal properties of kiwifruit
(Actinidia chinensis). Postharvest Biology and Technology, 20,
175e184.
Shenk, J. S., & Westerhaus, M. O. (1996). Calibration the ISI way. In
A. M. C. Davies, & P. C. Williams (Eds.), Near infrared
spectroscopy: The future waves (pp. 198e202). Chichester: NIR
Publications.
Staunton, S., Donald, D., & Pope, G. (2011). Estimating sugarcane
composition using ternary growth relationships. Proceedings of
the Australian Society of Sugar Cane Technologists, 33.
Torrecilla, J. S., Otero, L., & Sanz, P. D. (2004). A neural network
approach for thermal/pressure food processing. Journal of Food
Engineering, 62, 89e95.
Walsh, K. B., Guthrie, J. A., & Burney, J. W. (2000). Application of
commercially available, low-cost, miniaturised NIR
spectrometers to the assessment of the sugar content of intact
fruit. Australian Journal of Plant Physiology, 27, 1175e1186.
Wendte, K. W., Skotnikov, A., & Thomas, K. K. (2001). Sugar cane
yield monitor. U.S. Patent No. 6,272,819.
Williams, P. C., & Norris, K. H. (2001). Near-infrared technology in the
agricultural and food industries (2nd ed.). St. Paul, MN: AACC.
Wu, D., Feng, L., Zhang, C., & He, Y. (2008). Early detection of
Botrytis cinerea on eggplant leaves based on visible and near-
infrared spectroscopy. Transactions of the ASABE, 51(3),
1133e1139.
Xiaobo, Z., Jiewen, Z., Xingyi, H., & Yanxiao, L. (2007). Use of FT-
NIR spectrometry in non-invasive measurements of soluble
solid contents SSC of Fuji apple based on different PLS
models. Chemometrics and Intelligent Laboratory Systems, 87,
43e51.
Xing, J., Symons, S., Shahin, M., & Hatcher, D. (2010). Detection of
sprout damage in Canada Western Red Spring wheat with
multiple wavebands using visible/near-infrared hyperspectral
imaging. Biosystems Engineering, 106, 188e194.
b i os y s t e ms e ng i ne e r i ng 1 1 5 ( 2 0 1 3 ) 1 5 4 e1 6 1 161