基于近红外漫反射光谱及不同算法对泰来绿豆的产地溯源研究

为实现泰来绿豆产地溯源研究,筛选适宜算法与样品形态对模型精度的影响与应用,实验采用傅里叶变换近红外漫反射光谱技术对来自泰来绿豆和非泰来绿豆共计253 份样品进行产地溯源研究。结果表明,采用绿豆粉末状态建模效果要优于籽粒状态,对绿豆粉末鉴别研究中,运用因子化法建模的检测精度要优于标准算法,该法结合一阶导数+5点平滑的预处理方式建立的定性分析模型及采用Ward’s algorithm聚类算法结合二阶导数＋SNV＋5点平滑的预处理方式建立的聚类分析模型分别对粉末状泰来绿豆的正确鉴别率为96.15%和92.30%。对绿豆籽粒鉴别研究中,采用PLS法建立的定量分析模型,其RMSECV为0.129,R2为98.06,RPD为7.18,RMSEP为0.123,对泰来绿豆正确识别率为88.46%,要明显优于其他模型。因此,采用近红外漫反射光谱技术对泰来绿豆进行产地判别具有可行性。

Research on Origin Traceability of Tailai Mung Bean Based on Near Infrared Diffuse Reflectance Spectroscopy and Different Algorithms

In order to realize the study of origin traceability of Tailai mung bean, the influence and application of suitable algorithm and sample shape on model accuracy were screened. The origin of 253 samples from Tailai mung bean and non-Tailai mung bean was studied by Fourier near infrared diffuse reflectance spectroscopy. The results show that the modeling effect of mung bean powder state is better than that of grain state, and the detection accuracy of modeling by factoring method is better than that of standard algorithm. The qualitative analysis model based on the first derivative 5-point smoothing method and the Ward"s algorithm clustering algorithm combined with the second derivative + SNV+ 5-point smoothing method are used to establish the qualitative analysis model. The correct identification rates of Tailai mung bean were 96.15% and 92.30% respectively. The quantitative analysis model was established by the PLS method. The RMSECV was 0.129, R_² was 98.06, RPD was 7.18, the RMSEP was 0.123, and R_² was 98.01, and the correct recognition rate of the grain-like Tailai mung bean was 88. 46%, which was obviously superior to other models. Therefore, it is feasible to identify the origin of Tailai mung bean by near infrared diffuse reflectance spectroscopy.