[目的]研究不同风格及产地的卷烟.[方法]样品原始近红外光谱分别经过最小-最大归一化、一阶导数、一阶导数+矢量归一化、一阶导数+最小-最大归一化、二阶导数和连续小波变换(CWT)等方法处理后进行了主成分分析(PCA),建立了以马氏距离为基础的卷烟识别模型.[结果]小波变换模型有更好的样品识别能力,模型校正集样品的识别率为96.9%,检验集样品识别率为100%.分类可视化结果表明样品对本类型卷烟风格特征的表达最突出,国内烤烟型卷烟的国外混合型特征最不明显,国外混合型卷烟的国内烤烟型特征最不明显,国外烤烟型卷烟的国外混合型特征最不明显.[结论]采用近红外光谱技术可对不同风格及产地的卷烟进行正确识别.%[ Objective ] The research aimed to study different style characteristics and producing areas of cigarettes. [ Method ] The original spectra of samples were processed by minimum-maximum normalization, first-order derivative, first-order derivative + vector normalization,first-order derivative + minimum-maximum normalization, second-order derivative and continuous wavelet transform(CWT) individually. The processed spectra were treated by principal component analysis and the recognition models based on Mahalanobis distancewere built. [ Result]The recognition capability of CWT model was higher. The recognition rate of calibration set of CWT model was 96.9% and the recognition rates of validation set was 100%. The visualization results showed that all samples showed their own specific style characteristics best. It was also found that Chinese flue-cured cigarettes showed least characteristics of overseas blended cigarettes. Overseas blended cigarettes showed least characteristics of Chinese flue-cured cigarettes. Overseas flue-cured cigarettes showed least characteristics of overseas blended cigarettes.[ Conclusion ] It was proved that different style characteristics and producing areas of cigarettes could be correctly recognized by near infrared spectroscopy technology.
展开▼