多组分溶液近红外光谱检测算法研究

Research on the Detection Algorithm of Multi- component Solution Near Infrared Spectrum

利用下一代红外光谱检测仪MEMS-FTIR,在基于近红外波段1 000—2 100 nm区域对多组分糖溶液进行检测,通过开源平台的R语言对实验中的近红外光谱数据进行数据分析和PLS算法研究,为下一代快速、便携式、移动平台光谱数据分析奠定基础。本次研究主要使用具有开源性质的R语言和近红外光谱PLS算法,PLS算法可以显著提高近红外光谱回归模型的有效性。在建立回归模型后,对模型的RMSEP系数和R2系数进行分析和比较,结果表明建立的多组分糖浓度的PLS回归模型拟合程度较高。

This paper uses the next generation NIR spectrometer（ MEMS- FTIR） to detect the multi- component sugar concentration,based on the near- infrared wavelength from 1000 nm to 2100 nm. R language and PLS algorithm are applied to the NIR spectrum experiment to make regression modeling and data analysis; thus the experiment sets a foundation for the next generation of rapid,portable,multi- platform spectral data analysis. In the process of regression modeling and data analysis,the paper uses the R language which is open- source and the NIR PLS algorithm,and the PLS algorithm could significantly improve the effectiveness of NIR regression modeling. After regression modeling,the model of RMSEP and R2 coefficient are analyzed and compared,and the result of analysis shows that the PLS regression model of multi- component sugar solution has the better fitting effect.