探究了应用激光诱导击穿光谱(LIBS)结合化学计量学方法鉴别硫熏浙贝母的可行性。 采集了未经硫熏、 轻度硫熏和重度硫熏的浙贝母样本的LIBS光谱, 使用小波变换和归一化对原始光谱进行预处理后, 分别建立了基于全波段(400.41～871.65 nm)和基于特征波段(400.41～600.02 nm)的支持向量机(SVM)、 极限学习机(ELM)和随机森林(RF)的鉴别模型。 结果表明: 基于特征波段建立的三类模型的性能均与基于全波段建立的模型保持一致或更优, 说明特征波段的提取是有效的。 基于特征波段建立的模型中, SVM模型性能最优, 建模准确率和预测准确率分别达到了100%和95.83%。 综上所述, 应用LIBS技术结合特征波段提取和化学计量学方法鉴别不同程度硫熏的浙贝母是可行的, 且具有快速、 简便、 多元素同时分析的优势, 可为鉴别硫熏中药材提供依据, 有助于建立中药材质量检测与分级评定系统。

The paper discusses the feasibility of identifying Fritillaria thunbergii treated by sulfur fumigation using laser-induced breakdown spectroscopy (LIBS) technology with the combination of chemometric methods. Spectral data of Fritillaria thunbergii samples of no sulfur fumigation (no SF), mild sulfur fumigation (mild SF) and severe sulfur fumigation (severe SF) were collected and preprocessed by wavelet transform and normalization. Discrimination models of support vector machine (SVM), extreme learning machine (ELM) and random forest (RF) were developed on full spectra (400.41～871.65 nm) and characteristic wavebands (400.41～600.02 nm), respectively. Results showed that three models developed on characteristic wavebands all obtained the same or even better performance than the corresponding models developed on full spectra, indicating the effectiveness of extracting characteristic wavebands. Among the models developed on characteristic wavebands, SVM model obtained the optimal performance, with calibration and prediction accuracy reaching 100% and 95.83% respectively. The overall results demonstrated that LIBS technology with a combination of characteristic wavebands extraction and chemometric methods could be used for identifying Fritillaria thunbergii treated by sulfur fumigation. This study provides an instruction for identifying traditional Chinese medicine and can help to establish a quality detecting and grading evaluating system for traditional Chinese medicine.