基于近红外光谱技术快速测定甘薯多糖含量

摘 要：目的 建立一种基于近红外光谱技术快速测定甘薯多糖的方法。方法 通过采集来自不同地区的74个甘薯及甘薯干的近红外光谱图，对异常样本进行剔除与回收后随机选择其中56种作为校正集，11种作为验证集。通过一阶导数、二阶导数、多元散射校正（multiplicative signal correction，MSC）、标准正态变量变换（standard normal variate，SNV）等组合预处理方式对原始光谱进行处理，比较多元线性回归（stepwise multiple linear regression，SMLR）、主成分回归（principal component regression，PCR）和偏最小二乘法（partial least squares，PLS）三种方法建立的模型结果，进一步选择波段确定最佳甘薯多糖含量测定方法。结果 PLS建立的模型整体精确度和稳定性最佳，最优模型的预处理方式为一阶导数处理，该模型的最佳波段为全波段范围，校正集均方根误差（root mean square error of calibration set，RMSEC）为相关系数0.496，校正集相关系数（calibration set correlation coefficient，RC2）为0.9683，验证集均方根误差（root mean square error of prediction set，RMSEP）为0.430，验证集相关系数（prediction set coefficient of determination，RP2）为0.9440，主成分数为8。结论 通过近红外光谱技术结合偏最小二乘法建立甘薯多糖模型可作为甘薯多糖快速测定的可行性方法。

Rapid determination of sweet potato polysaccharide content based on near-infrared spectroscopy

Objective To establish a method for the rapid determination of sweet potato polysaccharides based on near-infrared spectroscopy. Methods The near-infrared spectra of 74 sweet potatoes and dried sweet potatoes from different regions were collected, after the abnormal samples were eliminated and recovered, 56 of them were randomly selected as the calibration set and 11 as the prediction set. The original spectrum was processed through a combination of preprocessing methods such as first derivative, second derivative, multiplicative signal correction (MSC), and standard normal variate (SNV). The results of the models established by the stepwise multiple linear regression (SMLR), principal component regression (PCR) and partial least squares (PLS) methods were compared, and the optimal method for determining the content of sweet potato polysaccharides was further selected by band selection. Results The model established by PLS had the best overall accuracy. The preprocessing method of the optimal model established by PLS was first-order derivative processing, and the optimal band of the model was in the range of 4000-10000 cm-1, the root mean square error of calibration set (RMSEC) was 0.50, correlation coefficient of calibration set (RC) was 0.9683, and the root mean square error in prediction (RMSEP) was 0.43, correlation coefficient of prediction set (RP) was 0.9440, and the number of principal components was 8. Conclusion The establishment of a sweet potato polysaccharide model by near-infrared spectroscopy combined with the partial least squares method can be used as a feasible method for the rapid determination of sweet potato polysaccharides.