[Effect of Optical Length on Detection Accuracy of Camellia Oil Adulteration by Near Infrared Spectroscopy].

In this research, near infrared spectroscopy was used to detect adulterated percent of camellia oil adulterated with soybean oil quantificationally at different optical lengths, and the effect of optical length on detection accuracy of adulterated percent was investigated. Soybean oil was put into camellia oil according to different mass fraction, the adulterated mass fraction was ranged from 1% to 50%. Transmission spectra of samples were acquired by a Quality Specspectrometer at different optical lengths (1, 2, 4, 10 mm), and effect of optical length on detection accuracy of adulterated percent was analyzed by comparing quantitative prediction models that developed at different calibration methods, pretreatment methods and wavelength range. The results indicate that the performance of quantitative prediction model of adulterated percent is improved as the optical length is increasing from 1 to 4 mm, while the performance of quantitative prediction model of adulterated percent is deteriorated as the optical length is increasing from 4 to 10 mm. 4 mm is a better optical length for camellia oil adulteration. The coefficients of determination of prediction (R2(P)) and root mean square error of prediction (RMSEP) in quantitative prediction models of adulterated percent for optical lengths of 1, 2, 4, 10 mm are 0.923, 0.977, 0.989, 0.962 and 4.58%, 2.54%, 1.72%, 3.20%, respectively.

[Discrimination of pressed and extracted camellia oils by Vis/NIR spectra combined with UVE-PLS-LDA].

Camellia oil is a special and high quality edible oil in China, and quality of pressed camellia oils is superior to extracted camellia oils. The objective of the present research was to discriminate pressed and extracted camellia oils by visible/near infrared (Vis/NIR) spectroscopy. The transmission spectra of pressed and extracted camellia oil samples were acquired using a QualitySpec spectrometer in the wavelength range of 350-1800 nm. Uninformative variable elimination (UVE) was used to select informative wavelength variables, and eliminate uninformative wavelength variables, then partial least squares combined with linear discriminant analysis (PLS-LDA) was used to develop classification model. At last, the classification model was used to discriminate 26 samples in the prediction set. The results indicate that UVE-PLS-LDA is an efficient discrimination and classification method, pressed and extracted camellia oils can be discriminated well by the classification model developed by UVE-PLS-LDA, the accurate rate is 100% for both samples in the calibration and prediction sets. So, Vis/NIR spectra combined with UVE-PLS-LDA is an effective method for discriminating pressed and extracted camellia oils.