论文部分内容阅读
基于支持张量机(STM)的三维荧光导数光谱定量分析方法,检测了食品中黄曲霉素。在计算三维荧光导数光谱时,将常规的、只适用于向量光谱数据的Savitzky-Golay方法扩展到由二阶张量描述的三维荧光光谱中。同时,应用了STM方法建立校正模型,对白酒和牛奶中的黄曲霉素进行了检测。在对白酒中的黄曲霉素检测中,复相关系数(CC)和预测误差均方根(RMSEP)分别为0.952 3和14.847 5,与常规的偏最小二乘(PLS)和支持向量机(SVM)方法相比,CC分别提高了2.40%和2.34%,RMSEP分别降低了8.92%和4.36%。在对牛奶中的黄曲霉素检测中,CC和RMSEP分别为0.996 5和5.448 9,与PLS和SVM的方法相比,RMSEP分别提高了0.40%和0.30%,RMSEP分别降低了18.31%和17.18%。检测结果表明,基于STM方法建立的校正模型要优于传统的SVM方法和PLS方法。
Based on the support vector machine (STM) three-dimensional fluorescence derivative spectroscopy quantitative analysis of aflatoxins in food. In calculating the three-dimensional fluorescence derivative spectrum, the conventional Savitzky-Golay method, which applies only to the spectral data of the vector, is extended to the three-dimensional fluorescence spectrum described by the second order tensor. At the same time, the STM method was applied to establish a calibration model for the detection of aflatoxins in liquor and milk. In the detection of aflatoxins in liquor, the complex correlation coefficient (CC) and root mean square of prediction error (RMSEP) were 0.952 3 and 14.847 5, respectively, which were in good agreement with the conventional partial least squares (PLS) and support vector machines SVM) method, CC increased by 2.40% and 2.34%, RMSEP decreased by 8.92% and 4.36% respectively. In the aflatoxins assay for milk, CC and RMSEP were 0.996 5 and 5.448 9, respectively, RMSEP increased by 0.40% and 0.30%, respectively, and RMSEP decreased by 18.31% and 17.18 respectively compared with PLS and SVM %. The test results show that the calibration model based on STM method is superior to the traditional SVM method and PLS method.