基于近红外光谱的茶叶高氯酸盐污染水平研究

Research on perchlorate contamination levels in tea based on near-infrared spectroscopy

中文关键词:  近红外光谱  茶叶  高氯酸盐  偏最小二乘判别分析  竞争性自适应重加权算法

英文关键词:near-infrared spectroscopy  tea  perchlorate  partial least squares-discriminant analysis  competitive adaptive reweighted sampling

中文摘要:

      目的 建立一种基于近红外光谱(near-infrared spectroscopy, NIR)分析技术的快速定量茶叶中高氯酸盐污染水平的预测模型。方法 采集不同产地、不同品种的91份茶叶干样, 通过傅里叶变换NIR扫描获得样品的近红外漫反射光谱, 使用超高效液相色谱-串联质谱法测定茶叶样品中的高氯酸盐含量, 以参考限量0.75 mg/kg为标准将样品分为两类; 利用偏最小二乘分析(partial least squares, PLS)建立高氯酸盐含量范围的预测模型, 同时使用一阶导(1st)、连续小波变换(continuous wavelet transform, CWT)、多元散射校正(multiplicative scatter correction, MSC)、标准正态变换(standard normal variate, SNV)等光谱预处理技术和竞争自适应重加权(competitive adaptive reweighted sampling, CARS)采样波长筛选技术对判别模型进行优化, 最后通过预测集样品对模型进行验证。结果 使用原始光谱建立的模型能够初步实现对高氯酸盐含量范围的预测, 而使用光谱预处理扣除光谱中的背景信息, 结合MSC和CARS方法共同处理后, 模型的预测正确度显著改善, 误判样品下降至3个, 预测正确率提高至88.5%。结论 本方法表明NIR技术可以为茶叶中高氯酸盐污染水平分析提供一种快速分析的新策略, 对茶叶产业高质量发展具有重要的实际意义。

英文摘要:

      Objective To develop a prediction model for the rapid quantification of perchlorate contamination levels in tea based on near-infrared spectroscopy (NIR) analysis technique. Methods Ninety-one dry tea samples from different origin and varieties were collected and obtained their near-infrared diffuse reflectance spectra by Fourier transform NIR scanning. The content of perchlorate in tea samples was determined by ultra performance liquid chromatography-tandem mass spectrometry. Samples were divided into two categories with reference limit of 0.75 mg/kg. The partial least squares (PLS) technique was used to establish a prediction model for perchlorate content ranges. Spectral preprocessing techniques such as first derivative (1st), continuous wavelet transform (CWT), multiplicative scatter correction (MSC), standard normal variate (SNV), and wavelength selection technique competitive adaptive reweighted sampling (CARS) were used to optimize the discriminant model, and the model was validated using prediction set samples. Results The model built using the original spectra could initially predict the perchlorate content range, while the prediction accuracy of the model was significantly improved by using the spectral pre-processing to deduct the background information in the spectra and combining the MSC and CARS methods together, and the misclassified samples were reduced to three and the prediction accuracy was improved to 88.5%. Conclusion The study indicates that NIR technology can be used as a new method for analyzing perchlorate contamination levels in tea, which has important practical significance for the high-quality development of the tea industry.