基于支持向量机的稻谷黄曲霉毒素B_1近红外无损检测

Near- infrared spectroscopy nondestructive determination of aflatoxin B_1 in paddy rice based on support vector machine regression

为提高近红外光谱技术检测贮藏稻谷中黄曲霉毒素B1准确性和精度,文章探讨以RBF径向基函数为核函数的支持向量机参数对建模精度影响。收集80个稻谷样本,应用近红外光谱采集光谱信息,采用化学分析方法测定贮藏稻谷中黄曲霉毒素B1含量,建立稻谷黄曲霉毒素B1支持向量机回归模型。确定基于RBF核函数模型最优参数为c=106,γ=0.0015。该模型校正集决定系数为0.913,校正标准偏差和预测标准偏差分别为1.186和1.267。结果表明,基于支持向量机算法建模可准确检测稻谷中黄曲霉毒素B1。为准确、实时监测贮藏稻谷黄曲霉毒素B1污染提供理论依据。

In order to improve the accuracy and the precision of the near infrared spectroscopy technical detection for the aflatoxin B1 in the paddy rice, this paper discussed the effects of the parameters of support vector machine on the modeling precision. The parameters took the RBF radial basis function as the kernel function. And 80 paddy rice samples were collected. The near infrared spectroscopy was used to collect spectral information. Then chemical analysis method was used to determine the content of aflatoxin B1 in the paddy rice. The regression model of support vector machine for aflatoxin B1 in the paddy rice was established. The optimal parameters based on the RBF radial basis function model were determined as c= 106 and y=0.0015. The correlation of calibration sets of this model was 0.913. The standard error of calibration sets and standard error of prediction were 1.186 and 1.267, respectively. The results showed that modeling based on the support vector machine algorithm could accurately detect aflatoxin B1 in the paddy rice. The research results could provide theoretical basis for accurate and real-time monitoring of aflatoxin B1 contamination in the storage paddy rice.