支持向量机—近红外光谱法用于真假奶粉的判别  

Applied Study on Support Vector Machines in Identifying Standard and Sub-standard Milk Powder with NIR Spectrometry

作　　者：吴静珠[1] 王一鸣[1] 张小超[2] 耿朝曦[1] 

机构地区：[1]中国农业大学 [2]中国农业机械化科学研究院,北京100083

出　　处：《农机化研究》2007年第1期155-158,共4页Journal of Agricultural Mechanization Research

基　　金：国家863高技术研究发展计划资助项目(2003AA209012)

摘　　要：将基于统计学理论的支持向量机(SVM)与近红外光谱分析技术相结合,对真假奶粉进行分类判别。以50个奶粉样品作为实验材料,通过SVM建立识别真假奶粉的模型。实验中采用高斯径向基函数(RBF)为核函数,根据SVM的不同输入量调整核参数γ建立最佳SVM模型,对学习机的38个样品识别率可达到100%,对预测集12个奶粉样品预测率可达到100%。实验表明,应用支持向量机—近红外光谱法建立判别真假奶粉的近红外定性分析模型,为真假奶粉的判别提供一个方便快捷的分析方法。Support vector machines （SVM） and the Fourier transform near infrared spectrometry （NIRS） have firstly been combined to build a classifier to identify standard and sub：standard milk powder. The radial basis ftmction is adopted as a kernel function of SVM. The effect of RBF parameter; which is adjusted according to the different input of SVM, is investigated. The training set is composed of：38 samples and the testing set is composed of 12samples. The correct classification ratio of the training set is up to 100%, while that of the testing set is up to 100%. The research results show that the combination of SVM and NIRS can be used as a fast and convenient tool to identify standard and sub-standard milk powder.

关 键 词：畜牧学 支持向量机 试验 近红外 奶粉 径向基函数 

分 类 号：S818.9[农业科学—畜牧学] O657.33[农业科学—畜牧兽医]