论文部分内容阅读
针对不同型号的近红外光谱仪器(主机:SupNIR-2700,从机:Nicolet AntarisⅡ)间的模型传递和同一仪器(Nicolet AntarisⅡ)不同分辨率的光谱间的模型传递进行研究,提出了一种改进的PDS算法-SPSG1st-PDS算法,该方法结合三次样条插值、Savitaky-Golay一阶求导和PDS算法。思路是通过三次样条插值拟在不破坏原始光谱固有的信息的前提下实现了主光谱与从光谱之间的匹配,然后对光谱进行S-G一阶求导去除光谱的基线漂移,再通过PDS算法进行模型传递,有效消除主、从光谱之间的差异,提高多元校正模型的预测精度。该方法用于醋酸乙烯酯在乙烯-乙酸乙烯酯共聚物中含量的研究,并且与小波去噪方法和S-G平滑方法作比较。实验表明:对于不同型号的仪器间的模型传递,新方法采用S-G一阶求导较其他方法有明显的优势,其验证集预测精密度RMSEP从20.595 0降低至0.374 8,明显优于S-G平滑(0.522 1)和小波去噪(0.516 7)方法,预测偏差也同样地被改善。对于同一仪器不同分辨率的光谱之间的模型传递,在模型传递前后其模型预测精密度RMSEP从0.272 2减少至0.255 3。通过提出的SP-SG1st-PDS算法,模型传递能应用于不同类型仪器之间,也能用于相同仪器不同分辨率的光谱之间,并且取得了满意的传递结果。
Aiming at the model transfer between different types of near-infrared spectrometer (host: SupNIR-2700, slave: Nicolet Antaris II) and the spectra transmission with different resolutions of the same instrument (Nicolet Antaris II), an improved PDS algorithm -SPSG1st-PDS algorithm, which combines cubic spline interpolation, Savitaky-Golay first-order derivative and PDS algorithm. The idea is to achieve the matching between the main spectrum and the spectrum by using the cubic spline interpolation without destroying the inherent information of the original spectrum, and then perform the SG first-order derivation on the spectrum to remove the baseline drift of the spectrum and then use the PDS algorithm Model transfer, effectively eliminate the main, from the difference between the spectrum, improve the multivariate calibration model prediction accuracy. This method is used to study the content of vinyl acetate in ethylene-vinyl acetate copolymer and compared with wavelet denoising method and S-G smoothing method. Experiments show that the new method using SG first-order derivation has obvious advantages over other methods for the model transfer among different types of instruments. The RMSEPEP of the verification set reduces from 20.595 0 to 0.374 8, which is obviously better than that of SG smoothing 0.522 1) and the wavelet denoising (0.516 7) method, the prediction bias is similarly improved. For model transfer between different resolution spectra of the same instrument, the model prediction precision RMSEP decreased from 0.272 2 to 0.255 3 before and after the model was transferred. Through the proposed SP-SG1st-PDS algorithm, the model transfer can be applied between different types of instruments as well as the spectra of different resolutions of the same instrument, and satisfactory transfer results have been achieved.