Nondestructive measurement of soluble solids content in pineapple fruit using short wavelength near infrared (SW-NIR) spectroscopy.
Suhandy, Diding
Introduction
The quality of pineapple fruit is measured not only by external
factors such as color, shape and size but also by internal factors such
as soluble solids content (SSC) and acidity. The SSC is considered to be
the most important factor in determining the quality of pineapple fruit.
The SSC is usually measured by using a refractometer but it is
time-consuming, destructive and not free waste.
In the recent years, the use of near infrared (NIR) spectroscopy
method both in the long wavelength and short wavelength for
nondestructive quality evaluation of some fruits has been reported. In
the long wavelength (1100-2500 nm), the use of this method for fruit
quality evaluation is limited due to high moisture content of fruit
(Krivoshiev, et al., 2000). In the short wavelength (700-1100 nm), the
NIR spectroscopy method successfully determined the quality parameters
of some fruits such as dry matter (DM), soluble solids content (SSC) and
sugar content, nondestructively.
NIR spectroscopy successfully measured the DM of some fruits such
as avocadoes (Clark et al., 2003), mango (Saranwong et al., 2003;
Suhandy et al., 2007a) and kiwi fruit (McGlone and Kawano, 1998). NIR
spectroscopy also successfully determined the SSC of some fruits such as
apple (Lammertyn et al., 1998; Park et al., 2003), melon (Dull et al.,
1992), mango (Suhandy et al., 2007b) and tomato (Slaughter et al., 1996;
Khuriyati and Matsuoka, 2004). The sugar contents of apple (Lu et al.,
2000), and orange (Kawano et al., 1993; McGlone et al., 2003) are also
successfully measured by NIR spectroscopy method.
In this research, the potentiality of SW-NIR spectroscopy to
measure the SSC of pineapple fruit was nondestructively evaluated. The
correlation between the near infrared spectra in the short wavelength
(700-1100 nm) and the SSC of pineapple fruit will be investigated. Then,
a calibration model for nondestructive SSC determination in pineapple
fruit using SW-NIR spectroscopy will be developed.
Materials and Method
Materials
A number of 42 pineapple fruits were hand harvested at the same
time from the same orchard and used as samples. To get a broad range of
SSC values, the samples consisting of three stages of maturation (i.e.
50% mature, 75% mature and 100% mature) were used. Then, the samples
were divided into two groups, calibration and validation set. Table 1
shows the sample characteristics of pineapple used in this research.
Method of spectral acquisition
Spectral acquisitions for each sample were taken at six different
positions using the NIR spectrometer (VIS-NIR USB4000; Ocean Optics,
USA). The spectra were stored in a computer for further analysis through
the fiber optics. Since temperature affects fruit spectrum, the fruit
temperature was maintained at 25[degrees]C by placing the sample into a
water bath for 30 minutes. The measuring condition for spectral
acquisitions was 150 ms for integration time and 100 scans for number of
scanning. A ceramic plate (diffuse reflectance standard model WS-1,
Ocean Optics, USA) was used as a reference. The intensity of light
transmitted through the ceramic plate was measured, and then NIR
measurement was performed by using a fruit in place of the ceramic
plate. Spectral acquisition of the ceramic plate was made every time
prior to the spectral acquisition of the fruit.
The absorbance spectra in the range of 300 nm to 1100 nm with 3 nm
intervals for each fruit were measured. Spectra in the short near
infrared region (700 nm to 1100 nm) were used for spectral analysis. The
absorbance spectra was obtained by using the following formula:
[A.sub.[lambda]] = -[log.sub.10]([S.sub.[lambda]]-
[D.sub.[lambda]]/[R.sub.[lambda]]-[D.sub.[lambda]]) (1)
where, [S.sub.[lambda]] = Intensity of sample at wavelength
[lambda] nm
[D.sub.[lambda]] = Intensity of dark at wavelength [lambda] nm
[R.sub.[lambda]] = Intensity of reference at wavelength [lambda] nm
Method of Soluble Solids Content (SSC) Measurement
The SSC of pineapple fruit was measured using refraktometer (Model
IPR 201, Atago, Japan). For this purpose, a portion of pineapple flesh
was cut at the point of spectral acquisition.
Data analysis
The average spectra from six positions were processed using
smoothing (number of segments: 5) and Savitzky-Golay second derivative
(left and right averaging: 33 nm, polynomial order: 2). Partial Least
Squares (PLS) regression was used to develop a calibration model. All of
these analyses were performed using The Unscrambler [R] version 7.01
(CAMO, Oslo, Norway), statistical software for multivariate calibration.
A student's t-test was performed using Statistical Package for the
Social Science (SPSS) version 11.0 for Windows in order to evaluate the
significance level of the model.
Results and Discussion
Analysis of pineapple fruit spectra
The pineapple fruit spectra were measured in absorbance mode at six
different positions. Then the original spectra for each sample were
transformed to its smoothing and second derivative spectra. Fig. 1
depicts the second derivative of pineapple fruit spectra in SW-NIR
region with high (16.11 Brix), middle (13.48 Brix) and low (11.99 Brix)
SSC values. As shown in Fig. 1, different spectra were identified due to
different SSC values.
[FIGURE 1 OMITTED]
Developing a Calibration Model
Using the PLS regression method the calibration and validation was
performed for original, smoothing and second derivative spectra (Table
2). Calibration model using the PLS method should have enough number of
latent variable (LV) to optimize the prediction model and to avoid
over-fitting. Furthermore, the best calibration model can be
characterized as follows. These are low number of latent variable (LV),
high coefficient of determination ([R.sup.2]), low standard error of
calibration (SEC), low standard error of prediction (SEP) and low bias.
The ratio of standard error of prediction to standard deviation (RPD)
value was the other parameter used for evaluating the performance of
calibration model. For good prediction model, it is clearly understood
that high RPD value is required (Williams, 1987).
Calibration model of original spectra at all wavelengths range
resulted in high coefficient of determination ([R.sup.2] = 0.89-0.97).
However, the calibration model for original spectra resulted in high
standard error of prediction (SEP = 1.02- 1.50). The number of latent
variable for original spectra was high for all wavelength range (LV =
6-10). For this reason it should be considered as a case of
over-fitting. In the second derivative spectra the number of latent
variable was too high (LV = 14-15) and too low (LV = 1). High
coefficient of determination was identified at some wavelengths range
([R.sup.2] = 0.96-0.98). However, those wavelengths range at the same
time have a very high SEP values (SEP = 1.33-1.57). It is also a case of
over-fitting.
In the smoothing spectra, the coefficient of determination was high
([R.sup.2] = 0.75-0.94). At the same time the SEP values were relatively
low (SEP = 0.88- 1.43). Then, the best calibration model was identified
at wavelength range of 700-960 nm for smoothing spectra with [R.sup.2] =
0.94 and SEC = 0.47. This wavelength range has relatively low factor (F
= 10) and low standard error of prediction (SEP = 0.88). For this
wavelength range the ratio of standard error of prediction to standard
deviation (RPD) value is also relatively high (RPD = 2.20).
In order to clarify the behavior of the calibration model, the
regression coefficient was plotted against the wavelength (Fig. 2). The
wavelength of 756 nm contributed to build the calibration model. This
wavelength corresponds with the absorbance band due to water
([H.sub.2]O) in the second overtone (Osborne et al., 1993). The
wavelength of 870 nm and 888 nm correspond with the absorbance band due
to carbohydrate in several form such as starch, sucrose, fructose and
glucose. These are the main component of SSC in fruit (Ho dan Hewitt,
1986). Khuriyati and Matsuoka (2004) used the wavelength of 884 nm for
SSC prediction in tomato. Saranwong et al. (2003) used the wavelength of
878 nm to determine the SSC of mango.
[FIGURE 2 OMITTED]
Validation of Calibration Model
The validation of calibration model resulted in low SEP and low
bias. Scatter plot between actual and predicted values is depicted in
Fig. 3. By a 95% confidence pair t-test, there were no significant
differences between the SSC of pineapple fruit measured using
refraktometer and that predicted by near infrared spectroscopy. This
result showed that a calibration model for nondestructive determination
of SSC in pineapple fruit using short wavelength near infrared (SW-NIR)
spectroscopy could be developed.
[FIGURE 3 OMITTED]
Conclusion
The nondestructive soluble solids content (SSC) measurement in
pineapple fruit using SW-NIR spectroscopy was successfully proposed. The
best calibration model was identified in the wavelength range of 700-960
nm for smoothing spectra with [R.sup.2] = 0.94 and SEC = 0.47. By a 95%
confidence pair t-test, there were no significant differences between
the SSC of pineapple fruit measured using refraktometer and that
predicted by near infrared spectroscopy. This result showed that a
calibration model for nondestructive determination of SSC in pineapple
fruit using SW-NIR spectroscopy could be developed.
References
[1] Clark, C.J., McGlone, V.A., Requejo, C., White, A., dan Woolf,
A.B. 2003. Dry matter determination in 'Hass' avocado by NIR
spectroscopy. Postharvest Biology and Technology. 29: 300-307.
[2] Dull, G.G., Leffler, R.G., Birth, G.S., dan Smittle, D.A. 1992.
Instrument for nondestructive measurement of soluble solids in honeydew
melons. J. Transaction of the ASAE 35: 735-737.
[3] Ho, L.C., dan Hewitt, J.D. 1986. Fruit development. In
"The Tomato Crop" (ed. By Atherton, J.G., Rudich, J.), Chapman
and Hall, New York, p. 206-239.
[4] Kawano, S., Fujiwara, K., dan Iwamoto, M. 1993. Nondestructive
determination of sugar content in Satsuma mandarin using near infrared
(NIR) transmittance. Journal of Japanese Society Horticultural Science.
62: 465-470.
[5] Khuriyati, N., dan Matsuoka, T. 2004. Near infrared
transmittance method for nondestructive determination of soluble solids
content in growing tomato fruits. Environ. Control in Biol. 42(3):
217-223.
[6] Krivoshiev, G.P., Chalucova, R.P., Moukarev, M.I. 2000. A
possibility for elimination of the interference from the peel in
nondestructive determination of the internal quality of fruit and
vegetables by VIS-NIR spectroscopy. Lebensm.u. -Technol.33: 344-353.
[7] Lammertyn, J., Nicolai, B., Ooms, K., De Smedt, V., dan De
Baerdemaeker, J. 1998. Nondestructive measurement of acidity, soluble
solids, and firmness of Jonagold apples using NIR spectroscopy. J.
Transaction of the ASAE 41(4): 1086-1094.
[8] Lu, R., Guyer, D.E., dan Beaudry, R.M. 2000. Determination of
firmness and sugar content of apple using NIR diffuse reflectance.
Journal of Texture Studies. 31(6): 615-630.
[9] McGlone, V.A., dan Kawano, S. 1998. Firmness, dry matter, and
soluble solids assessment of postharvest kiwifruit by NIR spectroscopy.
Postharvest Biology and Technology. 13: 131-141.
[10] McGlone, V.A., Fraser, D.G., Jordan, R.B., dan Kunnemeyer, R.
2003. Internal quality assessment of mandarin fruit by Vis/NIR
spectroscopy. J. Near Infrared Spectrosc. 11: 323-332.
[11] Osborne, B.G., Fearn, T., Hindle, P.H. 1993. Practical NIR
Spectroscopy with Application in Food and Beverage Analysis. Longman
Scientific and Technical, Harlow, UK, p 13-35.
[12] Park, B., Abbott, J.A., Lee, K.J., Choi, C.H., dan Choi, K.H.
2003. Near infrared diffuse reflectance for quantitative and qualitative
measurement of soluble solids and firmness of delicious and gala apples.
J. Transaction of the ASAE 46(6): 1721-1731.
[13] Saranwong, S., Sornsrivichai, J., dan Kawano, S. 2003.
Performance of a portable near infrared instrument for Brix value
determination of intact mango fruit. J. Near Infrared Spectrosc. 11:
175-181.
[14] Slaughter, D.C., Barrett, D., dan Boersig, M. 1996.
Nondestructive determination of soluble solids in tomatoes using near
infrared spectroscopy. Journal of Food Science. 61: 695-697.
[15] Suhandy, D., Hartanto, R., Prabawati, S., Yulianingsih.,
Yatmin. 2007a. Determination of dry matter in intact mango fruit using
near infrared spectroscopy/Original Paper in Indonesian Society of
Agricultural Engineering (ISAE) Conference.
[16] Suhandy, D., Hartanto, R., Prabawati, S., Yulianingsih.,
Yatmin. 2007b. Determination of the best NIR experimental condition for
nondestructive soluble solids content prediction in mango using near
infrared spectroscopy/ Original Paper in Agricultural Engineering
Journal/Bogor Agriculture University.21(4): 1-10.
[17] Williams, P.C. 1987. Variable affecting near infrared
reflectance spectroscopic analysis. In "Near-infrared technology in
the agriculture and food industries" (ed. by Williams, P. and
Norris, K.). Am. Soc. of Cereal Chemists Inc., St. Paul Minn., p
143-167.
Diding Suhandy
Laboratory of Bioprocess and Postharvest Engineering Department of
Agricultural Engineering Faculty of Agriculture, Lampung University Jl.
Soemantri Brojonegoro No.1 Bandar Lampung, Lampung Indonesia 35145
Corresponding author E-mail: diding2004@yahoo.com
Table 1: Statistical characteristics of sample used for
developing calibration and validation model for SSC
determination.
Items Calibration Set Validation Set
Number of sample 24 18
Range 11.25-18.44 11.87-17.89
Mean 14.76 14.74
Standard Deviation 1.85 1.94
Units %Brix %Brix
Table 2: Calibration and validation results for SSC of pineapple fruit
Type of Wavelength Latent [R.sup.2] SEC SEP
Spectra (nm) Variable
700-950 7 0.90 0.58 1.03
700-960 6 0.89 0.62 1.03
Original 700-970 7 0.93 0.49 1.02
Spectra 700-980 9 0.96 0.37 1.10
700-990 10 0.97 0.33 1.04
700-1000 9 0.94 0.45 1.50
700-950 15 0.98 0.28 1.55
2nd 700-960 1 0.22 1.63 1.78
Derivative 700-970 14 0.96 0.36 1.48
Spectra 700-980 15 0.97 0.32 1.33
700-990 14 0.97 0.31 1.57
700-1000 1 0.22 1.64 1.80
700-950 10 0.93 0.49 0.93
700-960 10 0.94 0.47 0.88
Smoothing 700-970 7 0.93 0.49 1.07
Spectra 700-980 9 0.94 0.43 1.29
700-990 6 0.75 0.92 1.36
700-1000 6 0.75 0.92 1.43
Difference
Type of between Bias RPD
Spectra SEC and SEP
0.45 0.51 1.88
0.41 0.46 1.89
Original 0.53 0.21 1.90
Spectra 0.73 0.35 1.76
0.71 0.38 1.86
1.05 0.43 1.30
1.27 0.06 1.25
--
0.15 0.02 1.09
2nd --
Derivative 1.12 0.02 1.31
Spectra --
1.01 0.29 1.46
1.26 0.02 1.23
--
0.16 0.01 1.08
0.44 0.51 2.07
0.41 0.40 2.20
Smoothing 0.58 0.38 1.81
Spectra 0.86 0.32 1.5
0.44 0.73 1.43
0.51 0.64 1.36
