基于2 种原理近红外光谱仪对辐照花生的快速鉴别比较

为比较不同近红外光谱仪在对辐照花生鉴别方面的差异，采用傅里叶变换、线性渐变分光2 种原理的近红外光谱仪针对不同辐照剂量的花生采集了近红外光谱数据，结合一阶导数、标准正态变量变换预处理，对0、1、2 kGy辐照剂量的花生建立偏最小二乘-判别分析模型，采用外部验证集样本对模型正确率进行验证。结果显示，线性渐变分光-近红外光谱仪的模型校正集正确率皆大于92%，交互验证集正确率皆大于85%，外部验证集正确率皆大于83%，普遍高于傅里叶变换-近红外光谱仪模型的正确率。因此，可得出结论：预处理对模型正确率的影响相对较小，不同仪器对模型正确率影响相对较大。

Comparison of Rapid Identification of Irradiated Peanuts Based on Two Near Infrared (NIR) Spectrometers

In order to compare the difference in the identification of 60Co γ-ray irradiated peanuts by different near-infrared
(NIR) spectrometers, NIR spectra of peanuts exposed to different dosages of irradiation were collected by Fourier transform
(FT) spectrometer and linear variable filter (LVF) spectrometer, respectively. The spectra were pretreated by the first
derivative and the standard normal variable (SNV), respectively. The partial least square-discriminat analysis (PLS-DA) calibration
models were developed and validated using external validation sets. The results indicated that the accuracy of the LVF-NIR model for
calibration, cross validation and external validation sets were larger than 92%, larger than 85% and larger than 83% respectively,
which were all larger than those of the FT-NIR model. It could be concluded that different data preprocessing methods have less
influence on the model accuracy, while the influence of different NIR spectrometers is significant.