基于投影寻踪法的烟碱近红外光谱定量模型研究
Development of near-infrared spectroscopy based quantitative model for nicotine with projection pursuit method
-
摘要:
为准确快速测量烟叶中的烟碱含量（质量分数），基于投影寻踪法建立烟叶中烟碱的近红外光谱定量模型。采用手持式近红外光谱仪和实验室近红外光谱仪分别获取校正集烟叶样本的近红外光谱和烟碱含量化学值；对光谱数据进行一阶求导后分别采用投影寻踪法和竞争性自适应重加权算法进行波长变量的选择，然后运用偏最小二乘法建立相应的烟碱含量定量预测模型。利用两种模型分别对3个批次的验证集样本进行预测，结果表明，基于投影寻踪法建立的烟叶中烟碱定量模型对烟碱含量的预测结果具有更小的平均绝对误差和更高的相关系数。该方法可为准确预测烟叶中的烟碱含量提供支持。 Abstract:In order to accurately and rapidly determine the nicotine content (mass fraction) in tobacco leaves, a quantitative model based on near-infrared (NIR) spectroscopy for nicotine content in tobacco leaves was established with projection pursuit method. The NIR spectra and nicotine contents of tobacco samples in calibration set were obtained by a handheld NIR spectrometer and a laboratory NIR spectrometer respectively. After the first-order derivation of spectral data, the wavelength variables were selected by projection pursuit method and competitive self-adaptive reweighted algorithm, and the corresponding quantitative prediction models for nicotine content were developed via partial least square method. The two developed models were used to predict the nicotine content in the samples of validation sets of three batches. The results showed that the nicotine contents in tobacco leaves predicted by the model developed with projection pursuit method had smaller mean absolute error and higher correlation coefficient. This method provides a support for the accurate prediction of nicotine content in tobacco leaves.
下载: