目的 探究不同贮存期温郁金颜色、气味和光谱信息变化，结合机器学习实现不同贮存期温郁金Curcuma wenyujin的快速区分；结合化学计量学，推测其特征差异成分。方法 采用多传感器结合判别函数分析、判别因子分析（discriminant factor analysis，DFA），软独立建模聚类分析（soft independent modeling of class analogies，SIMCA）、变量投影重要性指标（variable importance in projection，VIP）、相关性聚类标记热图分析、机器学习等方法，对不同贮存期温郁金的颜色、气味和光谱进行数据融合分析。结果 基于颜色信息建立了温郁金不同贮存期的判别函数，确定其颜色参数的范围。通过对气味信息的分类和筛选，表征出丙烯醛、丙酸、丙烯酸癸酯等9个特征差异成分。此外，“颜色-气味-光谱”数据融合，建立了决策树（decision tree，DT）、集成学习（ensemble learning，EL）分类模型，DT和EL模型的总准确度分别为85.40%和93.80%；经测试后的总准确度分别为91.7%和100.00%。结论 智能感官技术和近红外光谱技术可应用于贮存期的快速判别，为温郁金贮存期的快速区分与质量控制提供了新思路。

Objective This study investigated variations in color, odor, and spectral information of Curcuma wenyujin samples across different storage periods, aiming to achieve rapid discrimination via machine learning and to infer characteristic differential components through chemometric analysis. Methods Multi-sensor data fusion of color, odor, and spectral data was performed using discriminant function analysis (DFA), soft independent modeling of class analogies (SIMCA), variable importance in projection (VIP), correlation clustering with heatmap annotation, and machine learning algorithms. Results A discriminant function based on color information was established to classify different storage periods, and the corresponding parameter ranges were defined. Through screening of odor profiles, nine characteristic differential components, including acrolein, propionic acid, and decyl acrylate, were tentatively inferred. Furthermore, decision tree (DT) and ensemble learning (EL) classification models were developed based on “color-odor-spectrum” data fusion. They achieved overall accuracies of 85.40% and 93.80%, respectively. The overall accuracies after testing were 91.7% and 100.00%, respectively. Conclusion Intelligent sensory technologies combined with near-infrared spectroscopy were demonstrated to be applicable for rapid discrimination of storage periods, providing a novel strategy for the quality control of Curcuma wenyujin during storage.

R282.6

国家自然科学基金项目（82404862）；浙江省登峰学科（温州医科大学药学）（437502548）；温州医科大学人才科研启动项目（QTJ23033）；温州医科大学本专科学生科研立项资助课题（wyx2025101234，wyx2025101235）；浙江佐力药业股份有限公司横向课题项目（KJHX2009）