15 Applications: Food Science
351
matter, fat and sodium chloride, could be done with accuracies suitable for routine
analysis. RPD values of 6.0, 3.2 and 2.9 were obtained, respectively.
If is often difficult to develop calibration models in industry, due to the lack of
variation between the samples. The compound to be measured would cover only a
limited range. Filho and Volery [35] demonstrated how this can be overcome when
they quantified the solids content using a ‘broad-based’ calibration including five
different fresh cheeses with low, medium and high solids contents.
More recently, qualitative calibration model development has progressed considerably. González-Martín et al. [9] illustrated the power of NIR spectroscopy to predict
sensory attributes of cheese. Texture measurements such as hardness, chewiness and
creamy could be predicted with RPD values of 3.3, 2.7 and 1.6, respectively, with
the hardness measurements suitable for routine analysis. Taste predictions resulted
in RPD values of 1.6, 2.1, 2.3 and 2.6 for salty, buttery, rancid flavour and pungency,
respectively. Volatile compounds could also be measured with reasonable accuracy,
i.e. 2-nonanone (RPD = 3.4), acetaldehyde (RPD = 2.3), ethanol (RPD = 2.8),
2-heptanone (RPD = 2.8), 2-butanol (RPD = 2.1) and 2-pentanone (RPD = 2.0).
One of the most common methods of milk adulteration is the addition of water.
Adulteration with melamine which is harmful when consumed is, however, of much
greater concern. Melamine gives a false indication of increased protein content. The
difficulty in using NIR spectroscopy as a method of analysis [36] is the low levels of
melamine required to be detected.
15.6 Vegetable and Olive Oils
Sato et al. [37] performed the first NIR spectroscopy study on fats and oils. They
suggested that a spectral library could be compiled which could then be used to check
if the spectrum of an unknown sample matches any of the spectra in the database.
They continued with a study in which they successfully distinguished between nine
different types of vegetable oils, using principal component analysis (PCA) [37]. At
the same time, Bewig et al. [38] illustrated the use of discriminant analysis and only
four wavelengths to classify four different oils (cottonseed, peanut, soybean, canola).
Similarly, Hourant et al. [39] demonstrated the use of selected wavelength ranges
(1700–1800 and 2100–2400 nm) to classify seven vegetable oils.
The high value of extra virgin olive oil resulted in its potential adulteration with less costly oils [40]. Adulteration with inferior olive oils tends to be
of concern, especially in olive oil producing countries. In contrast, addition of
vegetable or seed oils seems to be of concern more likely in countries which
produce these oils and import olive oils. The most important indicator of adulteration is the fatty acid composition of the oil [41]. Detection and quantification of the type of adulterant in virgin olive oil at an accuracy of 75% were
demonstrated in an early study [42]. Using discriminant analysis, the authors
subsequently correctly identified the type of adulterant in extra virgin oil with a
351
matter, fat and sodium chloride, could be done with accuracies suitable for routine
analysis. RPD values of 6.0, 3.2 and 2.9 were obtained, respectively.
If is often difficult to develop calibration models in industry, due to the lack of
variation between the samples. The compound to be measured would cover only a
limited range. Filho and Volery [35] demonstrated how this can be overcome when
they quantified the solids content using a ‘broad-based’ calibration including five
different fresh cheeses with low, medium and high solids contents.
More recently, qualitative calibration model development has progressed considerably. González-Martín et al. [9] illustrated the power of NIR spectroscopy to predict
sensory attributes of cheese. Texture measurements such as hardness, chewiness and
creamy could be predicted with RPD values of 3.3, 2.7 and 1.6, respectively, with
the hardness measurements suitable for routine analysis. Taste predictions resulted
in RPD values of 1.6, 2.1, 2.3 and 2.6 for salty, buttery, rancid flavour and pungency,
respectively. Volatile compounds could also be measured with reasonable accuracy,
i.e. 2-nonanone (RPD = 3.4), acetaldehyde (RPD = 2.3), ethanol (RPD = 2.8),
2-heptanone (RPD = 2.8), 2-butanol (RPD = 2.1) and 2-pentanone (RPD = 2.0).
One of the most common methods of milk adulteration is the addition of water.
Adulteration with melamine which is harmful when consumed is, however, of much
greater concern. Melamine gives a false indication of increased protein content. The
difficulty in using NIR spectroscopy as a method of analysis [36] is the low levels of
melamine required to be detected.
15.6 Vegetable and Olive Oils
Sato et al. [37] performed the first NIR spectroscopy study on fats and oils. They
suggested that a spectral library could be compiled which could then be used to check
if the spectrum of an unknown sample matches any of the spectra in the database.
They continued with a study in which they successfully distinguished between nine
different types of vegetable oils, using principal component analysis (PCA) [37]. At
the same time, Bewig et al. [38] illustrated the use of discriminant analysis and only
four wavelengths to classify four different oils (cottonseed, peanut, soybean, canola).
Similarly, Hourant et al. [39] demonstrated the use of selected wavelength ranges
(1700–1800 and 2100–2400 nm) to classify seven vegetable oils.
The high value of extra virgin olive oil resulted in its potential adulteration with less costly oils [40]. Adulteration with inferior olive oils tends to be
of concern, especially in olive oil producing countries. In contrast, addition of
vegetable or seed oils seems to be of concern more likely in countries which
produce these oils and import olive oils. The most important indicator of adulteration is the fatty acid composition of the oil [41]. Detection and quantification of the type of adulterant in virgin olive oil at an accuracy of 75% were
demonstrated in an early study [42]. Using discriminant analysis, the authors
subsequently correctly identified the type of adulterant in extra virgin oil with a