脐橙总糖近红外光谱模型传递研究

摘要

In order to achieve NIRS （near infrared reflectance spectroscopy） calibration model sharing among similar NIRS instruments, an optimal NIRS partial least square （PLS） regression model was established using the major instrument and transfer to the minor instrument by either slope bias correction （SBC） or direct standardization （DS）. In addition, the dependence of calibration transfer on the number of standardization samples was explored. The results showed that the root mean squared errors of prediction （RMSEP） for the calibration model were 0.448% and 0.756% after standardization by DS and SBC, respectively. Thus, the DS method was better than the SBC method and could be used to successfully transfer the calibration model%为实现近红外光谱模型在同类光谱仪器间的共享,以脐橙总糖为例,在主仪器上建立最优的偏最小二乘（PLS）近红外光谱模型,采用斜率截距（Slope/Bias）校正法和直接校正（DS）算法把主仪器上建立的模型传递到从仪器上,并探讨标准化样品个数对模型传递效果的影响。研究表明：在主仪器建立的模型经DS算法传递后,从仪器预测集的预测标准差（RMSEP）为0.448%,经Slope/Bias校正法传递后的预测标准差（RMSEP）为0.756%,DS算法更优于Slope/Bias校正法,利用直接校正法实现脐橙总糖傅里叶近红外光谱模型传递是可行的。