摘要
应用近红外光谱分析技术(NIRS)并结合线性判别,对中老年和婴幼儿2种不同年龄阶段奶粉品种进行鉴别。对奶粉样本的原始光谱进行一阶微分处理后,采用Kennard-Stone法对80个样本进行挑选,选出60个作为训练集,剩余的20个作为预测集。采用留一交叉验证评价算法,选取最优主成分数2建立模型,并对测试集进行预测,正确率为100%。说明提出的方法具有很好的分类和鉴别作用,为婴幼儿奶粉的品种快速无损鉴别提供了一种新方法。
In this paper,a new method for discrimination of varieties of infant milk powder and middle-aged milk powder by means of near infrared spectroscopy (NIRS) technique combined with linear discriminant analysis (LDA) was developed.After first derivative of the original spectral of all milk powder,60 training set(total 80 samples) was selected using Kennard-Stone method,model,and the remainder samples formed the prediction set.2 was the best principal component with leave-one-out cross-validation algorithm.With the founded model,the recognition accuracy of 100% was achieved.This model is reliable and practicable.So the present paper could offer a new approach to the fast discrimination of varieties of milk powder.
出处
《乳业科学与技术》
2010年第3期125-127,130,共4页
JOURNAL OF DAIRY SCIENCE AND TECHNOLOGY
基金
2008年湖南省标准化战略资助项目
关键词
近红外光谱
线性判别
奶粉品种
鉴别
NIRS
linear discriminant Aanalysis
milk powder
discrimination