论文部分内容阅读
为考察烤烟烟碱近红外光谱定量模型在不同省份烤烟归属判别中的适用性,以贵州烤烟样品建立了烟碱的近红外定量模型,并通过对不同产区烤烟样品进行的主成分分析、马氏距离判断及模型预测结果相对误差分析确定了贵州模型的适用性.结果表明:①云南昭通、云南楚雄、云南保山、四川德昌的全部烤烟样品及湖南郴州的部分烤烟样品可适用于贵州模型,相对误差较小,主要在-5%~5%之间;福建三明、河南宝丰以及广东梅州的烤烟样品不适用于贵州模型,相对误差较大,大部分样品超出-5%~5%的范围.②将西南云贵川地区烤烟样品合并建模并与贵州模型进行比较,发现合并模型优于贵州模型,两种模型的校正均方根误差(RMSEC)分别为0.072 6和0.079 9;合并模型对云南四川样品和贵州样品的预测能力优于贵州模型,两种模型对烤烟样品的预测均方根误差(RMSEP)分别为0.076 0、0.079 9和0.109 0、0.084 6.合并模型增强了对云南、四川、贵州烤烟样品的预测能力,该研究结果为提高烤烟烟碱近红外光谱定量模型的适用性提供了参考.“,”In order to investigate the adaptability of NIR spectroscopy-based quantitative nicotine model to distinguish flue-cured tobacco leaves from different growing areas, an NIR quantitative nicotine model in flue-cured tobacco was established by taking Guizhou tobacco leaves as input samples. The adaptability of the established Guizhou model was determined via principal component analysis (PCA), Mahalanobis distance analysis and predicted data relative error analysis using tobacco samples from different areas. The results indicated that: 1) The Guizhou model could be adapted to all leaf samples from Zhaotong, Chuxiong, Baoshan in Yunnan and Dechang in Sichuan and some leaf samples from Chenzhou in Hunan with relative error from -5% to 5% , while it did not adapt well to the samples from Sanming in Fujian, Baofeng in Henan and Meizhou in Guangdong with large relative error (higher than 5% or lower than -5%). 2) A merged model was established by using the samples from Yunnan, Sichuan and Guizhou and was compared with the Guizhou model. The results indicated that the merged model was superior to the Guizhou model. The RMSECs of the merged and Guizhou models were 0.072 6 and 0.079 9, respectively. The predictive ability of the merged model to Yunnan, Sichuan and Guizhou samples was better than that of the Guizhou model, the RMSEP of the merged model was 0.076 0 and 0.079 9 for Yunnan and Sichuan samples and Guizhou samples respectively, and that of the Guizhou model was 0.109 0 and 0.084 6,respectively. The merged model improved the predictive ability to the samples from Yunnan, Sichuan and Guizhou. The results of this study provide references for improving the adaptability of NIR quantitative model for the nicotine in flue-cured tobacco.