段式正交信号校正方法及在小麦近红外光谱数据分析中的应用

针对光谱数据峰宽、局部效应显著、含有噪音、变量个数多及彼此间常存在严重的复共线性等问题,改进和设计一种光谱数据局部校正方法:基于窗口平滑的段式正交信号校正方法,并将之结合偏最小二乘回归,以实现光谱数据的预处理及定量分析。通过NIPALS算法初始化将滤去的正交成分,以近邻分段方式进行逐个波长点的正交信号校正。而后将去噪后的光谱矩阵作为新的自变量阵,通过偏最小二乘回归构建其与性质参变量间的校正模型。通过小麦近红外漫反射光谱数据的应用实验结果表明,本方法正交成分估计稳定,去噪明显,模型的预报性能优于其它方法,PLS成分数减少,模型更加简洁。

Piecewise Orthogonal Signal Correction Approach and Its Application in the Analysis of Wheat Near-infrared Spectroscopic Data

Taking into account the local effect sensitivity and numerous predictor variables with serious multicollinearity of spectra data,an improved piecewise orthogonal signal correction(POSC)approach was constructed based on the nonlinear iterative partial least squares(NIPALS)algorithm.Then a novel partial least squares(PLS)method that embedded the improved POSC into the PLS regression framework,termed as improved POSC-PLS method,was implemented.The improved POSC technique was firstly used by selecting a set of spectra with a special moving window to pretreat the spectra matrix and eliminate the local variance,thus the pretreated spectra matrix was taken as the new independent variables matrix,then the PLS algorithm was applied to build the calibration model.Finally,application of the proposed POSC-PLS approach to the wheat NIR diffuse reflectance spectra quantitative analysis was presented for comparison with the MLR(multiple linear regression),PLS,OSC-PLS and old POSC-PLS.The result indicates that the improved POSC-PLS approach not only can ensure the pretreated spectra matrix robust,but also improves the model prediction accuracy and decreases the PLS factors than models obtained by the above methods and so its resulting model becomes more concise.So,preprocessing with the improved POSC is shown to benefit the multivariate PLS model because it performs a localized regression modeling procedure.The improved POSC is a potential chemometric technique for the pretreatment of various spectra.