用多维校正法提高近红外牛奶成分校正模型稳健性的研究

乳品质量问题是制约乳品工业发展的瓶颈。近红外光谱技术以其快速、简便等优越性广泛应用于各个领域，在乳品检测中有巨大的应用潜力和前景。但样品温度对近红外分析模型的稳健性和检测结果会产生较大影响，为获得较好的分析准确度，文章提出了抑制温度影响的新方法——多维正交校正(NOSC)结合多维偏最小二乘(NPLS)法。它将温度作为建模元素组成含温度信息的三维光谱矩阵，先将三维光谱阵进行预处理，运用NOSC方法滤除无关信息的干扰，再采用NPLS方法建立牛奶成分与透射光谱的校正模型。文章在25，30，35和40 ℃四个温度点处做了研究，波长范围为1 100～1 700 nm。通过常规法和多维校正法来减小温度的影响。实验结果表明，常规法无法抑制温度的影响，而与其他方法相比，NOSC-NPLS法建立的校正模型具有更好的预测能力和稳健性，能更有效地消除温度的影响。

Research on the Robustness Improvement of Calibration Model for Measuring the Contents of Components in Milk by Multidimensional Calibration in Near-Infrared Spectroscopy

A new hybrid algorithm (NOSC-NPLS),the combination of multi-way orthogonal signal correction (N-OSC) algorithm and multi-way partial least squares (N-PLS) algorithm,is proposed. In NOSC-NPLS algorithm,the 3-D spectral matrix was firstly constructed,which is composed of the temperature information and the NIR spectrum. Secondly,the N-OSC algorithm was used as a pretreatment algorithm to remove the interference information irrelevant to analyte in 3-D spectral matrix. Finally,the N-PLS algorithm was applied t...