主成分分析结合Fisher最优分割法在烟叶分切中的应用

摘要

To rationalize leaf sectioning prior to leaf threshing, the cured leaf samples of grade C3F of cv. Hongda were sliced according to Golden section, the near infrared spectra of each section were processed by principal component analysis (PCA) to extract the principal components of near infrared spectra. The sectioning for tobacco leaf was studied by combining with Fisher optimal segmentation method (FOSM), and compared with the existing method in terms of chemical components and sensory quality. The results showed that: 1)The sample could be sliced into 11 sections according to Golden section, two principal components were determined via PCA, and their cumulative variance contribution rate was up to 99.789%. 2) Based on the comprehensive score of principal components and combining with FOSM, the optimal sectioning ratio for tobacco leaf of grade C3F of cv. Hongda was determined as base∶middle∶tip=33%∶29%∶38%. 3)The contents of total sugar, reducing sugar and nicotine in base, middle and tip differed significantly or extremely significantly, and their sensory scores were 44.87, 47.70 and 49.06, respectively. Meantime, the sensory scores of middle and tip promoted to different degrees comparing with those sectioned with the existing method. It provides a reference for scientific processing and reasonable use of flue-cured tobacco leaves.%为探索烟叶原料分段打叶复烤较科学、合理的烟叶分切方法，以红大C3F等级初烤烟叶为研究对象，采用黄金分割法将烟叶进行分切并采集各段近红外光谱，运用主成分分析提取各段近红外光谱主成分，结合Fisher最优分割法对烟叶分切方法进行了研究，并从内在化学成分和感官质量两方面与现有分切方法的分段效果进行了对比评价。结果表明：①采用黄金分割法能将红大C3F等级初烤烟叶分切成11段，通过对各段近红外光谱进行主成分分析，最终确定了两个主成分用于反映原料信息，其累积方差贡献率达99.789%。②以主成分综合得分为基础，结合Fisher最优分割法确定了红大C3F等级初烤烟叶在实际生产中的最佳分段比例为叶基∶叶中∶叶尖=33%∶29%∶38%。③采用Fisher最优分割法分切后烟叶总糖、还原糖和烟碱等指标含量在叶基、叶中、叶尖之间的差异达到显著或极显著水平，叶基（44.87）、叶中（47.70）、叶尖（49.06）感官质量得分差距较大，且叶尖和叶中的感官质量得分相比现有分切方法均有不同程度提升，这为烟叶分段后科学打叶复烤加工与合理利用提供了参考。