为减小不同比色皿对近红外光谱测量结果的影响,提高水质酸度定量分析模型的预测精度,探讨了正交信号校正(OSC)用于不同比色皿的光谱背景干扰的去除效果。用两个同一批次的石英比色皿对32个不同pH的水样装样,采集近红外光谱数据,采用OSC对原始光谱进行预处理。比较OSC预处理前后两组光谱间的差异,并建立了水质酸度的偏最小二乘(PLS)定量分析模型,分析了光谱差异对模型预测精度的影响。结果表明:经OSC预处理后,两组光谱的平均差异值由0.004 2降低至0.001 3,光谱校正率达90%;与原始光谱建立的PLS模型相比,基于OSC预处理后的光谱建立的PLS模型的预测精度显著提高,预测均方根误差差值由0.912降低至0.205,相关系数差值由0.364降低至7.00×10^-3,二者分别减小了78%和98%。

In order to reduce the effect of different cuvettes on the measurement results of near infrared spectroscopy, as well as improve the prediction accuracy of quantitative analysis model for the acidity of water, the removel effect of orthogonal signal correction (OSC) on the spectral background interference from different curvettes was discussed. Two cuvettes from the same batch were used to load 32 water samples with different pH values, and near infrared spectral data were collected, and OSC was used to preprocess the original spectra. The difference between the two groups of spectra before and after OSC preprocessing was compared, and the partial least squares (PLS) quantitative analysis models were established to analyze the effect of spectral difference on the prediction accuracy of the models. As shown by the results, after OSC preprocessing, the average difference value between the two groups of spectra decreased from 0.004 2 to 0.001 3, and the spectral correction rate reached 90%. Compared with the PLS models established by the original spectra, the prediction accuracy of the PLS models established by the spectra after the OSC preprocessing was significantly improved, and the difference values of predicted root-mean-square error decreased from 0.912 to 0.205, and that of correlation coefficient decreased from 0.364 to 7.00×10^-3, and both of them reduced by 78% and 98%, respectively.