Determination Of Main Quality Indices Of Green Tea By Near-infrared Reflectance Spectroscopy

Near-infrared spectroscopy(NIRS), with the characteristics of high speed, non-destructiveness, high precision, reliable detction data, online analysis, etc., is a pollution-free qualitative and quantitative analysis method, and has been applied in many fields. We just utilized NIRS to estimate the influences of different calculation methods, spectral data preprocessing methods and spectral wavelengths on the quantitative calibrations: moisture, free amino acids(AA), water-solubility protein, total tea polyphenols (TP), caffeine, and cellulose of green tea. The results were as follows:1.The influences of different calculation methods on the performance of quantitative calibrations. Comparing the effects of principal components regression (PCR) and partial least square (PLS) on quantitative calibrations, the results indicated that PCR was the optimal method for quantitative calibrations of moisture, and PLS was the optimal method for the rest 5 components.2. The influences of different spectral data preprocessing methods and spectral wavelengths on the performance of quantitative calibrations. After choosing the super calculation methods, the study selected appropriate spectral data preprocessing methods and spectral regions for the 6 components. Moisture: Smooth Savitzky-Golay 9 Points(sg-9), 4593.64-10001.1 cm^-1; water-solubility protein: First Derivative BCAP and Normalization by Closure*( db1 and nc1), 4439.36-9002.14cm^-1; caffeine: MSC full* and First Derivative BCAP(mf and db1), 4393.08-4801.91 and 5392.03-6603.11 and 7798.77-10001.1cm^-1;amino acids:none,4393.08-4801.91 and 5392.03-6603.11 and 7798.77-10001.1cm^-1; total tea polyphenols: First Derivative BCAP(db1), 4439.36--9002.14cm^-1; Cellulose: MSC full*(mf), 4393.08-4801.91 and 5392.03- -6603.11 and 7798.77-10001.2 cm^-1.3.Evaluation of the 6 quantitative calibrations. The calibrations can be rated via the "Quality Value". The Q-value takes all important statistical values(such as SEE, SEP, V-bias, etc.) into account. The higher of the Q-value and the better the calibration will be. The 6 quantitative Calibrations' Q-value are 0.9650(moisture), 0.9335(total tea polyphenols), 0.9604(water-solubility protein), 0.9372(caffeine), 0.8521(amino acids) and 0.8239(Cellulose). All the Q-value are above 0.75, which indicated that the 6 Calibrations could be used to determine these 6 components of tea.4. The preliminary application of the 6 quantitative Calibrations. 14 green tea samples that were neither in calibration nor in validation sets were verified to furtherly evaluate these quantitative calibrations. Coefficients of correlation(R) between NIRS predicted and actual of moisture(0.9948), total tea polyphenols(0.8965), water-solubility protein(0.9694), caffeine(0.8661), amino acids(0.9379) and cellulose(0.9695) were respectively obtained in these samples. Thus, the moisture, cellulose, amino acids, and water-solubility protein quantitative Calibrations' accuracy of prediction could be comparable to chemical methods. And the quantitative Calibrations of total tea polyphenols and caffeine need to be optimized.