ATR-FTIR结合RBF神经网络对市售西红花的产地溯源

Geographical Traceability of Traded Saffron (Crocus sativus L.) by ATR-FTIR and RBF Neural Network

目的:利用衰减全反射傅里叶变换红外光谱法测定市售西红花的红外谱图,对不同产地西红花进行判别。方法:对129个市售西红花样品红外光谱图进行采集,选择500~1800 cm-1区域的光谱图作为产地判别的数据来源。结果:应用主成分分析法对西红花样品光谱进行分析,可将西红花区分3大主要产地区域。西红花产地追溯模型是通过数据预处理后的ATR-FTIR光谱学与RBF神经网络建模,小波去噪预处理方法结合RBF神经网络模型对3大区域的判别正确率最高,预测集正确率为96.6%。结论:衰减全反射傅里叶变换红外光谱技术结合化学计量学方法是一种快速、有效的技术手段,可用于区分不同市售西红花样品的产地溯源。

Objective:Attenuated total reflectance fourier transform infrared spectroscopy(ATR-FTIR)combined with multivariate analysis has been applied for the discrimination of 129 traded saffron samples.Methods:Theinfrared spectra of saffron samples were recorded.The best discriminatory approach was achieved inthe spectral region 500~1800 cm-1.Results:Principal component analysis was applied to samples spectral regions.Saffron samples had be discriminated three main areas of geographical origin.The saffron origin traceability modelwas developed by preprocessed ATR-FTIR coupled with soft independent modeling of RBF neural network.From the original grouped cases,the correctly classified rate was 96.6%.Conclusion:The combination of infrared spectroscopic technique with multivariateanalysis is a rapid and effective method to discriminate tradedsaffron samples in terms of geographical origin.