352
M. Manley and P. J. Williams
90% accuracy [42]. The level of adulteration could also be accurately predicted
(±0.9% w/w).
Downey et al. [4] used SIMCA to classify authentic extra virgin olive oils from
the same oils adulterated with sunflower oil. It was possible to detect adulteration at
levels as low as 1% (w/w) as well as to predict the level of sunflower oil added using
PLS regression (SECV = 0.8% w/w). They developed a model that could determine
the level of sunflower oil adulterant with an accuracy suitable for industry use. Subsequent studies detected the adulteration of olive oils with a range of adulterants with
very low error limits [43]. A recent study confirmed the use of NIR spectroscopy
as a method to screen for adulterated olive oils [14]. When an unadulterated sample
was also analysed, the level of detection was as low as 2.7% (w/w). Using SIMCA
and without an unadulterated sample, the level of detection was less accurate (20%).
The use of handheld instruments has also been considered for oil analysis [44].
In spite of the handheld device only using the wavelength range of 950–1650 nm,
lard adulteration in palm oil could be detected with a model accuracy of more than
0.95 using SIMCA. Using PLS regression gave even better results (R
2
= 0.99). As
is the case with many adulteration studies, the sample set was limited, thus only
demonstrating the feasibility of the application.
15.7 Fruit and Vegetables
One of the earliest fruit-related studies, detection of adulteration of orange juices, was
reviewed by Shilton et al. [45]. As was the case with the early studies on oil, Shilton
et al. [45] suggested the use of NIR spectroscopy as a ‘fingerprint’ technique rather
than trying to predict specific constituent levels. In a subsequent study, however,
LDA and PLS were used to classify apple juices up to 100% correctly, based on fruit
variety [4].
The ability of NIR instruments, in association with chemometrics, to predict
fruit and vegetable quality properties has been comprehensively reviewed [46].
Studies considered include dry matter content of onions, soluble solids content
(SSC) of apples and water in mushrooms. Prediction of acidity was less accurate
than predicting SSC due to NIR spectroscopy not being able to measure it directly,
but based on its correlation with sugars. Similarly, fruit maturity could be predicted
based on its correlation with sugar content and the microstructure of the fruit tissue.
The microstructure of the fruit affects how the NIR light penetrates into and scatters
within the fruit tissue which enables measurement of stiffness, internal damage as
well as sensory attributes.
The successful measurement of changes in soluble solids and dry matter in individual mango fruit over time during ripening was demonstrated with a handheld
device (950–1650 nm) [47]. The penetration depth of about 7.4 mm into the fruit
tissue ensured representative sampling and contributed to the success of the developed
models.
M. Manley and P. J. Williams
90% accuracy [42]. The level of adulteration could also be accurately predicted
(±0.9% w/w).
Downey et al. [4] used SIMCA to classify authentic extra virgin olive oils from
the same oils adulterated with sunflower oil. It was possible to detect adulteration at
levels as low as 1% (w/w) as well as to predict the level of sunflower oil added using
PLS regression (SECV = 0.8% w/w). They developed a model that could determine
the level of sunflower oil adulterant with an accuracy suitable for industry use. Subsequent studies detected the adulteration of olive oils with a range of adulterants with
very low error limits [43]. A recent study confirmed the use of NIR spectroscopy
as a method to screen for adulterated olive oils [14]. When an unadulterated sample
was also analysed, the level of detection was as low as 2.7% (w/w). Using SIMCA
and without an unadulterated sample, the level of detection was less accurate (20%).
The use of handheld instruments has also been considered for oil analysis [44].
In spite of the handheld device only using the wavelength range of 950–1650 nm,
lard adulteration in palm oil could be detected with a model accuracy of more than
0.95 using SIMCA. Using PLS regression gave even better results (R
2
= 0.99). As
is the case with many adulteration studies, the sample set was limited, thus only
demonstrating the feasibility of the application.
15.7 Fruit and Vegetables
One of the earliest fruit-related studies, detection of adulteration of orange juices, was
reviewed by Shilton et al. [45]. As was the case with the early studies on oil, Shilton
et al. [45] suggested the use of NIR spectroscopy as a ‘fingerprint’ technique rather
than trying to predict specific constituent levels. In a subsequent study, however,
LDA and PLS were used to classify apple juices up to 100% correctly, based on fruit
variety [4].
The ability of NIR instruments, in association with chemometrics, to predict
fruit and vegetable quality properties has been comprehensively reviewed [46].
Studies considered include dry matter content of onions, soluble solids content
(SSC) of apples and water in mushrooms. Prediction of acidity was less accurate
than predicting SSC due to NIR spectroscopy not being able to measure it directly,
but based on its correlation with sugars. Similarly, fruit maturity could be predicted
based on its correlation with sugar content and the microstructure of the fruit tissue.
The microstructure of the fruit affects how the NIR light penetrates into and scatters
within the fruit tissue which enables measurement of stiffness, internal damage as
well as sensory attributes.
The successful measurement of changes in soluble solids and dry matter in individual mango fruit over time during ripening was demonstrated with a handheld
device (950–1650 nm) [47]. The penetration depth of about 7.4 mm into the fruit
tissue ensured representative sampling and contributed to the success of the developed
models.