论文部分内容阅读
模型标准化方法是目前最常用的近红外光谱模型传递方法,但大都基于单尺度,校正不够精细且易导致光谱信息丢失.本文尝试采用多尺度建模这一化学计量学领域刚兴起的新思路,将其应用于近红外光谱模型传递中,在多尺度空间上校正主、从机光谱之间的差异,有效实现了预处理与多元校正的一体化运算.将该算法应用于多台近红外光谱仪所采集的玉米模型转移中,结果表明,本算法无需预处理即可实现模型传递功能,传递后的模型校正精度和可靠性都得到了显著提升,进而为近红外光谱模型传递提供了一种新手段.
However, most of them are based on single-scale, and they are not precise enough and lead to the loss of spectral information easily.This paper attempts to use multiscale modeling, a new idea just emerging in the field of chemometrics, The proposed method is applied to the near infrared spectroscopy (NIRS) model transfer, and the differences between the master and slave spectra are corrected in multi-scale space to effectively realize the integration of preprocessing and multivariate calibration. The algorithm is applied to several NIRS The results show that the proposed algorithm can realize the model transfer function without preprocessing, and the accuracy and reliability of the model calibration after delivery have been significantly improved, thus providing a new method for the near infrared spectroscopy model transfer means.