近红外光谱结合PLS-DA算法对人发与动物毛发的鉴别

Identification of Human Hair and Animal Hair by Near Infrared Spectroscopy with PLS-DA Algorithm

采集了247份人发与247份动物毛发的近红外光谱图,对比了不同预处理方法和建模波段下所建模型的效果,选出了最优的预处理方法为一阶导数,最佳的建模波段为6000~5500 cm^(-1)+5000~4500 cm^(-1)。建立了人发与羊毛的PLS-DA判别模型,模型的决定系数R^(2)达到了0.9775,验证集判别正确率达到了100%。使用此模型可有效区分人发与动物毛发,为假发材质的鉴别提供了一种新的方法。

The nearinfraredspectra of 247 human hair and 247 animal hair were collected,and the effects of different pretreatment methods and modeling bands were compared.The fist derivative pretreatment method was selected as the optimal pretreatment method and 6000~5500 cm^(-1)+5000~4500 cm^(-1) was selected as the best waveband to modeling.The PLS-DA discriminant model of human hair and wool was established.The determination coefficient R^(2) of the model reached 0.9775,and the discriminant accuracy of verification set reached 100%.This model can effectively distinguish human hair from animal hair.It provides a new method for the identification of wig materials.