Karl Norris导数滤波器结合主成分分析转移NIR模型

Transfer of Near Infrared Calibration Models Combining with Karl Norris Derivative Filter and Principal Component Analysis

分别以Nicolet Antaris、Spectrum One NTS近红外光谱仪为源机和目标机,采集153个具有代表性的初烤烟叶样品光谱,应用Karl Norris导数滤波器对其中的120个烟叶样品的光谱进行预处理,筛选谱区,同时结合主成分分析验证源机与目标机光谱的一致性,将获取的有关光谱预处理参数应用于源机模型的转移运算得到目标机的转移模型,应用源机模型和目标机转移模型对另外的33个烟叶样品的总糖、还原糖、总氮和烟碱进行预测,并对预测结果进行配对t-检验。结果显示,在P>0.05时,二者的预测结果无显著性差异。

Of the 153 representative flue-cured tobacco samples scanned using Nicolet Antaris and PE Spectrum One NTS-NIR Spectrometer as source and target meters, respectively, the NIR spectral data of 120 samples were pretreated by Karl Norris derivative filter for spectral regions screening, and the spectral coincidence between source and target meters was proved with principal component analysis. The transferred models of the target meter were developed by applying the obtained pretreatment parameters into the transfer calculation of the source meter＇ s models. The total sugars, reducingsugars, total nitrogen and nicotine in 33 remaining tobacco samples were predicted with the transferred models of the target meter and the models of the source meter separately, and the obtained values were compared by t-test in pairs, the results indicated that there was no significant difference between the predictive values when P 〉 0. 05.