基于化学计量学和近红外光谱法的油莎豆产地溯源

Traceability of geographical origin of Cyperus esculentus based on chemometrics and near infrared spectroscopy

中文关键词:  油莎豆  近红外  K最近邻算法  产地溯源

英文关键词:Cyperus esculentus  near infrared spectroscopy  K-nearest neighbor algorithm  geographical origin

中文摘要:

      目的 采用近红外光谱技术对油莎豆进行分析, 并应用化学计量学中识别模式对油莎豆进行产地溯源。方法 采用近红外光谱法结合化学计量学软件, 对来自河北、湖南、山东、新疆、云南等地408份油莎豆样品进行产地溯源, 分别采用多元散射校正、多量标准化或多量标准化耦合去趋势算法3种光谱预处理方法和支持向量机(support vector machine, SVM)、簇类独立分类(soft independent modeling of class analogy, SIMCA)、正交偏最小二乘判别(orthogonal partial least squares discriminant analysis, OPLS-DA)、偏最小二乘判别(partial least squares discriminant analysis, PLS-DA)、和K最近邻算法(K-nearest neighbor algorithm, KNN)等5种识别模式进行产地识别。结果 SVM、SIMCA、OPLS-DA、PLS-DA和KNN等5种模式的建模识别率分别为91.89%、94.47%、62.37%、65.32%和100.00%。选择KNN作为产地识别模型, 分析不同预处理方法、数据预处理及样本距离对模型预测结果稳定性的影响, 筛选出最优模型参数。选用多元散射校正光谱预处理方式, 在UV标度化、Pareto标度化、自动标度化或中心化任一种数据预处理条件下, 样本距离选用街区距离, 测试集识别率能达到100.00%。结论 近红外光谱结合KNN模式的技术具有分析速度快、操作简单、样本预处理容易、无损、在线的定性定量分析等优点, 具有一定应用前景。

英文摘要:

      Objective To analyze Cyperus esculentus by near infrared spectroscopy, and trace geographical origin of Cyperus esculentus by the identification model in chemometrics. Methods A total of 408 samples of Cyperus esculentus samples from Hebei, Hunan, Shandong, Xinjiang, and Yunnan were analyzed for provenance tracing using near-infrared spectroscopy and chemometric software, 3 kinds of spectral preprocessing methods including multiplicative scatter correction, standard normal variate transformation and standard normal variate transformation & detrending, were used respectively, and 5 kinds of recognition modes such as support vector machine (SVM) , soft independent modeling of class analogy (SIMCA), orthogonal partial least squares discriminant analysis (OPLS-DA), partial least squares discriminant analysis (PLS-DA), and K-nearest neighbor algorithm (KNN) were used to identify the geographical origin. Results The modeling recognition rates of the 5 kinds of modes including SVM, SIMCA, OPLS-DA, PLS-DA, and KNN were 91.89%, 94.47%, 62.37%, 65.32%, and 100.00% respectively. The KNN was selected as the origin identification model, and the impact of different preprocessing methods, data preprocessing and sample distance on the stability of the model prediction results were analyzed in order to select the optimal model parameters. The prediction set recognition rate could reach 100.00% by using multiplicative scatter correction spectral preprocessing method, one of data preprocessing methods including UV, Pareto, automatic, or centering, and block distance as the sample distance. Conclusion The technology of near infrared spectroscopy combined with KNN mode has the advantages of fast analysis speed, simple operation, easy sample pretreatment, non-destructive, on-line qualitative and quantitative analysis, etc., and has a certain application prospect.