期刊论文

中文

食品工业科技

1002-0306

2012

33

4

89-91,96

10.13386/j.issn1002-0306.2012.04.014

国家科技支撑计划项目（2009BADB9B07）；

本校为第一机构

采用连续投影算法（successive project algorithm,SPA）对177个不同产地油菜蜜样本的近红外光谱做波长选择,然后以33个特征变量作线性识别分析（LDA）。同时,也采用了主成分分析（PCA）对变量进行压缩。比较了二次识别分析（QDA）和簇类独立软模式分类法（SIMCA）的鉴别结果。SPA-LDA模型预测集的鉴别准确率为97.7%,而PCA-LDA、全谱的SIMCA和SPA-QDA预测集的正确率分别为93.2%、95.4%和90.9%;上述四种方法ROC曲线下的面积分别为0.964、0.912、0.948和0.933。SPA-LDA性能比其他三种方法要好。该方法准确、可靠,为蜂蜜真实性的现场快速检测提供了一种新方法。

Near infrared spectroscopy combined with chemometrics methods were used to detect Brassica honey samples.For this purpose,successive project algorithm（SPA） was employed to choose thirty-three spectral variables,which were used as input of liner discriminative analysis（LDA） model.For comparison,principal components analysis（PCA） was used to compress the high-dimensional data.Quadratic discriminative analysis（QDA） and soft independent modeling of class analogy（SIMCA） were also employed to classify authentic and adulterated honey.The cla...