基于反向比例解析的近红外光谱定量模型快速构建方法

A Rapid Construction Method for Near Infrared Spectral Quantitative Model Based on Reverse Proportional Analysis

为解决光谱漂移问题,该研究设计了一种基于反向比例解析的近红外光谱定量模型方法。以烟叶近红外光谱和烟碱含量为研究对象,将数据划分为训练集和测试集。通过计算训练集光谱与测试集光谱的相关性并按照高低排序,选择前20%的光谱,运用约束规划的方法,计算测试集的拟合系数,得到测试集光谱的估计值。结果显示,使用反向比例解析法建立的模型的平均绝对误差为0.3466,预测标准偏差为0.4252,相关系数为0.7932,优于PLS模型。反向光谱比例解析可以有效解决光谱漂移问题,实现烟草中烟碱含量的准确预测,为烟碱的有效测量提供参考。

To address the issue of spectral drift,this study proposes a quantitative model approach for near infrared spectroscopy based on reverse proportional analytical method.The research focuses on tobacco leaf near infrared spectra and nicotine content,with the data being divided into training and test sets.By calculating the correlation between the spectra in these sets and sorting them accordingly,select the top 20%of spectra for constraint programming calculation to estimate the spectra in the test set.The results demonstrate that the reverse proportional analytical method yields an average absolute error of 0.3466,a predicted standard deviation of 0.4252,and a correlation coefficient of 0.7932,indicating its strong performance compared to PLS models.This highlights how the reverse spectral proportional analytical method effectively addresses spectral drift while accurately predicting nicotine content in tobacco,providing valuable insights for nicotine measurement.