基于NIR和PLS-DA法的东北大米产地快速溯源方法研究

鉴于市场上东北大米掺伪和假冒现象频发,本研究将近红外光谱技术与PLS-DA判别法相结合用于东北大米产地快速溯源。实验共收集75份大米样本(52份东北产地大米,23份非东北产地大米),按照4∶1的比例随机划分得到训练集样本60份和测试集15份。为消除品种、颗粒形态等对近红外光谱和产地鉴别的影响,实验比较多种光谱预处理方法,并根据相关系数值确定了矢量归一化为首选的预处理方法;将矢量归一化预处理后的光谱按照吸收峰分4个区间分别建立东北大米产地PLS-DA模型,其中4 000～5 500cm~(-1)谱区建立产地溯源模型,训练集准确率可达93.33%,测试集准确率为86.67%。实验结果表明,近红外光谱与PLS-DA法结合在东北大米产地快速溯源技术领域有着光明的应用前景。

Research on ChineseSNortheast Rice Origin Traceability Based on Near-Infrared Spectroscopy and Partial Least Square-Discriminant Analysis

In view of the frequent occurrence of adulteration and counterfeiting in Chinese Northeast rice, this work focus on the rapid origin traceability of Northeast rice combined near-infrared spectroscopy (NIR) and Partial Least Square-Discriminant Analysis (PLS-DA). 75 rice samples (52 Northeast producing areas and 23 non-Northeast producing areas) were collected and randomly divided into training set (60 samples)and test set(15 samples) in proportion of 4:1. In order to eliminate the influence of variety and particle morphology, several kinds of spectral pretreatment methods were applied and compared. Vector normalization were selected finally according to the correlation coefficient. To trace the Northeast Rice Origin rapidly , the Ssegmentation models were respectively established according to the 4 main absorbance regions based on the spectrum treated by vector normalization and PLS-DA. Among them, The model built in the 4000-5500 cm-1 spectral region had the better performance. The recognitionSrate of the training set reached to 93.33% and that of the test se is 86.67%.The experimental results show that the combination of NIR and PLS-DA has a bright application prospect in the field of rapid and non-destructive traceability of Northeast rice.