光谱分辨率的选取对食用油近红外模型性能影响分析

近红外光谱仪的类型和工作参数的设定对测量过程和分析结果均有一定影响.重点讨论了在傅立叶变换型光谱仪上设定不同的光谱分辨率对食用油近红外定量模型性能的影响.实验采用Vertex 70光谱仪,在3种光谱分辨率(4,8,16 cm-1)条件下,采用透射式液体光纤探头采集60份食用油样品近红外谱图.针对3组近红外光谱样品集,分别采用蒙特卡罗采样法剔除异常样品,根据Kennard-Stone法划分校正集和校验集后,建立优化食用油棕榈酸近红外定量分析模型并作预测.结果表明基于16 cm-1建立的食用油近红外模型指标优于4 cm-1和8 cm-1所建的模型,该结果可以为均匀液体作近红外检测时光谱分辨率的设定提供参考.

Research on NIR Model of Edible Oil Impacted by Spectral Resolution Determination

There are some effects on near infrared analysis result caused by the type of near infrared spectrometer and working parameters settings.How the different spectral resolution on Fourier transform spectrometer impact quantitative NIR model of edible oil was discussed in this paper.The experiment collected near infrared spectrum of 60 edible oil samples using Vertex 70 spectrometer,in three kinds of spectral resolution(4,8,16 cm-1) conditions,by the transmission type liquid optical fiber probe.Monte Carlo sampling method was used to analysis three groups of near infrared spectral sample set and eliminate the abnormal samples.The calibration set and validation set were divided according to Kennard-Stone method.The quantitative models of palm acid were established,optimized and validated.Experiments shown that the model built based on 16 cm-1 had better performance than 4 cm-1 and 8 cm-1.The results can provided a reference spectral resolution while detecting uniform liquid by near infrared technology.