基于近红外光谱与CARS-PLS-LDA的不同产地雪菊的定性鉴定

Qualitative Identification of Coreopsis tinctoria from Different Habitats by Near Infrared Spectroscopy and CARS - PLS - LDA

目的雪菊是新疆特有的药食两用物种,研究雪菊的近红外光谱快速定性鉴定方法对于该产品的开发利用具有重要的意义。方法采用近红外光谱结合竞争性自适应权重法-偏最小二乘-线性判别分析(CARS-PLS-LDA)对新疆不同产地的雪菊样品进行判别。利用CARS法筛选有效的波长分类变量,然后采用PLS-LDA法建立不同产地雪菊样品的分类鉴定方法,以敏感性、特异性、鉴定正确率作为最终的评价指标。结果CARS-PLS-LDA可以提取雪菊样本中有效的分类信息,实现6个不同产地的雪菊样本的有效分类鉴定;CARS-PLS-LDA法使用的波长变量数仅为全部变量的不到2.0%,所有种类雪菊预测集样本的敏感性、特异性、鉴定正确率均为100%,结果优于未经波长筛选的PLS-LDA分类方法。结论CARS-PLS-LDA法降低了鉴定方法的复杂度,同样获得了良好的鉴定准确率,近红外光谱与CARS-PLS-LDA相结合可以有效鉴定不同产地的雪菊样品。

Coreopsis tinctoria is a native medicinal and edible plant resources,so exploring its rapid qualitative identification will benefit to development and utilization of the resources.This research was conducted to attempt to discriminate Coreopsis tinctoria varieties by the method that near infrared spectroscopy combined with competitive adaptive reweighed sampling-partial least squares-linear discriminant analysis(CARS-PLS-LDA).CARS was applied to select effect wavelength information about clas-sification from full spectrum,and the classification method of Coreopsis tinctoria from various place were established by PLS-LDA.The evaluation index including sensitivity,specificity and accuracy were used to judge the final results.The research showed that the effect classification informations of the samples can be extracted by CARS-PLS-LDA,which was applied to classify Coreopsis tinctoria from six different habitats every well.Only 2.0%of the full spectrum was applied to identify,and the three evaluation index of samples from six different habitats in validation set were 100%.Fortunately,It results were better than the PLS-LDA method without CARS.The results indicated that the classification method is simpler with a good accuracy,and near infrared spectroscopy combined with CARS-PLS-LDA is allowed to effctively identify Coreopsis tinctoria from different hab-itats.