Prediction of Titratable Acid in Bayberry Juice by Near-Infrared Spectroscopy

摘要

Bayberry has a high economical and officinal value. In this paper, the use of near-infrared spectroscopy (NIRS) was explored as a tool to detect the titratable acid in bayberry juice. Calibration models were developed from GB 12293-90 in China as a reference method and NIRS data using partial least squares (PLS) regression and principal components regression (PCR). Different preprocessing methods and different wave bands were applied. The correlation coefficients (r) and root mean squares error of prediction (RMSEP) in the best model for titratable acid was 0.9623 and 2.85, with the range of 12500-5405 cm~(-1). The results indicate that it is very feasible to predict the quality of bayberry juice using NIRS technique. The prediction accuracies can not be improved by using preprocessing methods. Due to the short consuming time and low cost of monitoring, NIRS technique has its potential for the rapid, reliable and nondestructive prediction of titratable acid in bayberry juice.