本文采用时间分辨荧光发射谱（TRFES）结合化学计量学建立了一种准确、快速的山茶油产地识别方法。收集浙江、江西和湖南的山茶油样品共180个，采集它们的TRFES并从稳态荧光发射和荧光衰减维度对荧光信号的特点进行了对比；利用平行因子分析（PARAFAC）对训练集样品数据进行降维和特征优化；最终选取了两个因子的因子得分作为人工神经网络（ANN）的输入并建立山茶油产地识别模型。结果表明，相较于稳态荧光发射，荧光衰减受荧光分子浓度影响较小。因此TRFES被认为指纹性极强，有利于山茶油产地识别。山茶油产地识别模型的验证相关系数为98.7%，预测相关系数为96.1%，表明该模型稳健、准确，适合山茶油产地识别。最终表明，TRFES结合化学计量学分析可以完成对山茶油产地的准确、快速识别。

An accurate and rapid method for geographical identification of camellia oil was established based on the combination of time-resolved fluorescence emission spectrum (TRFES) and chemometrics. Totally 180 samples from Zhejiang, Jiangxi and Hunan were collected and their TRFES were compared from the dimensions of steady-state fluorescence emission and fluorescence decay; Parallel factor analysis (PARAFAC) was performed for the dimensional reduction and the optimization of characteristics based on training set; two factors were selected and then their factor scores were utilized as the input of artificial neural network (ANN), and finally the geographical identification model of camellia oil was established. The result showed that fluorescence decay was less influenced by the concentration of fluorophores than steady-state fluorescence emission. Hence, TRFES displayed stronger fingerprint characteristics and could be good for the geographical identification of camellia oil. The cross-validation coefficient was 98.7% and prediction coefficient was 96.1% for the geographical identification model of camellia oil, indicating the strong robustness and high accuracy of the model, which is considered to be suitable for the geographical identification of camellia oil. This study proved that geographical identification of camellia oil could be completed by the combination of TRFES and chemometrics.

国家自然科学基金面上项目（31271874）