近红外离散波长光谱基线漂移校正方法研究

微分是(近)红外光谱多元分析校正中最常使用也是最有效的光谱基线漂移校正方法。由于数据数目较少及相邻数据在光谱意义或数学意义上缺乏连续性，微分不能直接用于离散波长光谱消除基线漂移。为此，提出了一种结合插值拟合和微分校正离散光谱基线漂移的新方法。思路是采用三次样条插值法对离散波长光谱进行拟合，然后对拟合光谱进行Savitaky-Golay卷积求导，再从微分光谱中取出对应于原离散波长光谱数值的数值，构成离散波长光谱的微分光谱，从而实现离散波长光谱的基线漂移校正。通过分别由模拟离散波长光谱数据和实际的离散波长光谱数据建立多元校正模型检验新方法效果。采用ABC干粉灭火剂和土壤的近红外光谱数据及性质建立了PLS和MLR模型。结果表明，新方法能有效消除离散波长光谱的基线漂移对多元分析校正产生的不利影响，明显地提高了多元分析校正模型的准确性，对改善离散波长光谱仪器分析准确度具有重要的理论意义和实际应用价值。

A Correction Method of Baseline Drift of Discrete Spectrum of NIR

In the present paper, a new correction method of baseline drift of discrete spectrum is proposed by combination of cubic spline interpolation and first order derivative. A fitting spectrum is constructed by cubic spline interpolation, using the datum in discrete spectrum as interpolation nodes. The fitting spectrum is differentiable. First order derivative is applied to the fitting spectrum to calculate derivative spectrum. The spectral wavelengths which are the same as the original discrete spectrum were taken out from the derivative spectrum to constitute the first derivative spectra of the discrete spectra, thereby to correct the baseline drift of the discrete spectra. The effects of the new method were demonstrated by comparison of the performances of multivariate models built using original spectra, direct differential spectra and the spectra pretreated by the new method. The results show that negative effects on the performance of multivariate model caused by baseline drift of discrete spectra can be effectively eliminated by the new method.