| The quality of fresh tea shoots is the basis of the tea quality assurance, it is also the premise of the standardized processing to differentiate the classification of fresh tea shoots. At present, the quality analysis of fresh tea shoots is mainly done by sensory evaluation and chemical inspection. The sensory evaluation is of much subjective and require rich experience,the chemical analysis is of high cost and time consuming. Both can not be used for online analysis.Therefore, it is urgent to establish an accurate and fast method for fresh tea shoots’acquisition and processing.Near infrared spectroscopy (NIRS) has been widely used in petroleum, pharmaceutical, tobacco, beverage and food industry due to the advantages of fast, accurate, easy and non-destructive.This paper attempts to use near infrared (NIR) spectroscopy to determine main chemical compositions contents and evaluate the quality grade of Huang shan Maofeng tea shoots.The main results of the dissertation involve:(1) Three quantitative analysis models for tea shoots,including moisture,total nitrogen and crude fiber,were built by applying near infrared spectroscopy combined with partial squares (NIR-PLS).The effects on the spectral preprocessing methods and principal components factors (PCs) to results were discussed. The optimal calibration models were evaluated and the prediction performance was validated by independent validation sets. Experimental results showed that,the spectral preprocessing of the models were first derivative (1stDer)>1stDer added multiplication scatter correction (MSC)> Elimination constant offset, The optimal numbers of PLS factors were 9、8、 9. The determination coefficient (R2C) of internal cross-validation were 0.9414、 0.9528,0.9355,respectively. The root mean square errors of cross-validation (RMSECV) were 0.32,0.0883,0.0953, respectively. The determination coefficient (R2P) of external validation were 0.9109,0.8989,0.8895, respectively.The root mean square errors of prediction (RMSEP) were 0.361,0.103,0.195,respectively. The values of relative prediction deviation (RPD) were 4.14,3.61,3.94.The results demonstrated that it was feasible to apply NIR spectroscopy to determine rapidly main chemical compositions contents of tea shoots。(2) Class Correlation model based on three main contents by BP-ANN were built. Using different contents as the input layer, the output layer is the quality grade.Meanwhile,training algorithms, architecture, number of neurons in each layer, the number of layers were adjusted. Experimental results showed that, using three quantitative compositions contents including moisture, total nitrogen and crude fiber as the input layer,the hidden layer neurons number was 8,the output layer neurons number was 1, the determination coefficient (R2) of the validation was 0.9644, The results demonstrated that, the three composition contents were closely related to the tea shoots’quality grade,which laid a foundation for the building of tea quality grade model by NIR-PLS.(3) NIR speetroscopy combined with partial leas square (PLS) was employed for building the quantitative models of fresh tea shoots’ quality grade.Experimental results showed that in the spectra region between 13282.9 cm-1 and 12016.4 cm-1, 10190.4 cm-1 and 9134.9 cm-1,8638.9 cm-1 and 7287.9 cm-1,6844.6 cm-1 and 5831.3 cm-1, spectral preprocessing of the models was elimination constant offset, principal components was 10, the discrimination ratio was 93.10%.Experimental results showed that the model had high Prediction precision and indicated that NIR spectroscopy technique could be efficiently employed to evaluate fresh tea shoots’ quality grade. |
PDF Full Text Request