不法分子通常会在毒品中掺杂其他物质来增加毒品质量，获取高额利润。为了实现掺杂毒品快速准确的定性、定量分析，同时探究不同建模方法、不同波段光谱以及数据降维等对模型分类结果的影响，本文采用衰减全反射-傅里叶变换红外光谱仪采集了咖啡因、葡萄糖、扑热息痛、非那西汀和淀粉5种添加组分与盐酸甲基苯丙胺混合后的红外光谱数据（共计135份），对数据进行预处理后进行分类模型的构建。实验结果表明：特征变量与原始变量相比能达到更高的分类识别准确率；多层感知器和径向基函数可用于5种添加物的分类识别，但不适用于区分不同质量分数的甲基苯丙胺样品；对光谱指纹区数据使用因子分析降维结合Bayes判别分析（BDA）构建的分类模型，在16维和33维变量处能够分别实现5种添加物和不同质量分数甲基苯丙胺样品的完全区分，准确率可达100%。该研究实现了甲基苯丙胺与5种添加组分快速准确的定性定量分析，为毒品来源的推断提供了科学的数据支撑，也为基层民警处理此类案件提供了理论支撑和方法参考。

Criminals typically adulterate drugs with other substances to increase drug-quality to obtain high profits， which has adverse effects on the society. The purpose of this study is to achieve fast and accurate， qualitative and quantitative analysis of adulterated drugs， and explore the influence of various factors on the results of model classification， e.g.， the modeling method， the spectral band， and dimension reduction. Here， 135 infrared spectrograms of caffeine， glucose， acetaminophen， phenacetin， and starch mixed with methamphetamine hydrochloride of different mass fraction were collected using attenuated total reflection-Fourier transform infrared spectrometer. A classification model was then constructed after data preprocessing. The results demonstrate that the characteristic variable has higher classification accuracy than the original variable. Multilayer perceptron （MLP） and radical basis function （RBF） could be used to classify and identify five additives； however， they were not able to distinguish methamphetamine samples with different mass fractions. A classification model of spectral fingerprint data was constructed using factor analysis and dimension reduction combined with Bayes discriminant analysis （BDA）， and the complete differentiation of five additives and methamphetamine samples with different mass fraction was achieved on 16 and 33 dimensional variables， with accuracy of up to 100%. We achieved fast and accurate qualitative and quantitative analysis of methamphetamine and five added components， which provide scientific data support for the inference of drug source. The results also provide theoretical support and method reference for drug-related cases.