Coupled with partial least squares(PLS),near infrared(NIR)spectroscopy was applied to develop a fast and nondestructive method to identify the production date of Rizhao green tea aiming at the deficiencies of the exis...Coupled with partial least squares(PLS),near infrared(NIR)spectroscopy was applied to develop a fast and nondestructive method to identify the production date of Rizhao green tea aiming at the deficiencies of the existing methods.In the modeling process,the raw spectra were first processed by five-point smoothing and first derivative.And then,moving window back propagation artificial neural network(MW-BP-ANN)was applied to select the characteristic spectral variables.After that,the calibration model was built by PLS,and the optimum model was achieved when 9 principal component scores(PCs)were included.The performances of the calibration models were evaluated according to root mean square error of predictionεRMSEP,correlation coefficient(C p)and residual prediction deviation(σRPD).The optimum results of the calibration model was achieved,andεRMSEP=19.965,C p=0.943 andσRPD=3.07.The overall results sufficiently demonstrate that NIR spectroscopy combined with PLS can be efficiently applied in the rapid identification of green tea production date.展开更多
基金Supported by the State Key Laboratory of Sensor Technology Fund(SKT1202)China Postdoctoral Science Foundation(2012M521319)thecrosswise project "Application of micro NIR spectrograph in the wireless sensornetwork"(2015-1-1273)
基金National Basic Research Program of China(No.JSJL2016210A001)State Key Laboratory of Sensor Technology Fund(No.SKT1507)
文摘Coupled with partial least squares(PLS),near infrared(NIR)spectroscopy was applied to develop a fast and nondestructive method to identify the production date of Rizhao green tea aiming at the deficiencies of the existing methods.In the modeling process,the raw spectra were first processed by five-point smoothing and first derivative.And then,moving window back propagation artificial neural network(MW-BP-ANN)was applied to select the characteristic spectral variables.After that,the calibration model was built by PLS,and the optimum model was achieved when 9 principal component scores(PCs)were included.The performances of the calibration models were evaluated according to root mean square error of predictionεRMSEP,correlation coefficient(C p)and residual prediction deviation(σRPD).The optimum results of the calibration model was achieved,andεRMSEP=19.965,C p=0.943 andσRPD=3.07.The overall results sufficiently demonstrate that NIR spectroscopy combined with PLS can be efficiently applied in the rapid identification of green tea production date.
