近红外光谱的数据分析方法研究进展

Research Progress on Data Analysis Methods of Near Infrared Spectroscopy

数据预处理、特征选择和建立模型是近红外光谱分析技术中三个重要的过程。从降低噪声、消除基线漂移、校正散射光三方面介绍数据预处理方法;介绍基于区域的光谱特征选择方法和基于单变量的光谱特征选择方法;并根据应用不同,论述了定量分析和定性分析两类建模方法。现就这三个过程中常用方法的基本原理、优缺点进行综述,为相关研究者提供参考。

Data preprocessing,feature selection and model building are three important processes in the near infrared spectroscopy analysis technology.The data preprocessing method is introduced from three aspects:reducing noise,eliminating baseline drift and correcting scattered light.The spectral feature selection method based on region and the spectral feature selection method based on single variable are introduced.According to different applications,two modeling methods of quantitative analysis and qualitative analysis are discussed.This paper summarizes the commonly used method in the process of the three basic principle and disadvantages of each method,provide a reference for related researchers.