对牛奶成分近红外光谱分析方法的研究

研究牛奶成分近红外光谱测量的基本方法，探讨了在用偏最小二乘(PLS)方法建立系统校正模型过程中，采用直接正交(DO)数据预处理方法滤除牛奶漫反射光谱中与待测组分质量浓度无关的干扰信息的可行性，并与多元散射校正(MSC)及二阶微分(SOD)等光谱数据预处理方法进行比较．实验结果表明：相对于DO处理前，PLS模型的实际预测偏差(RMSEP)明显改善，最佳主成分数降低；相对于其他数据预处理方法，DO处理后系统PLS模型的RMSECV及RMSEP相对较低．DO数据预处理方法是近红外光谱分析中，从复杂重叠光谱中提取净信号信息、滤除噪音、进一步改善对原始光谱数据解释的有效途径．

Development of NIR Spectral Analysis Method for Milk Constituents

NIR Spectral Analysis Method for milk constituents was explored. The Direct Orthogo-nalization(DO) pre-processing method was discussed to filter the noise signal which is irrelevant to the concentration being measured to milk reflection spectra, the PLS method was employed to build correction model of the system by Cross Validation way. The predictions of the Partial Least Squares (PLS) Regression models with and without the DO pre-processing, with other pre-processing methods such as Multiplicative Scatter Correction (MSC) and Second Order Derivative(SOD) are evaluated. After the DO pre-processing, the Root Mean Square Error of Prediction(RMSEP) of the PLS model is reduced significantly, and the number of Principal Component(PC) is reduced. Compared to other methods, the Root Mean Square Error of Cross Validation(RMSECV) of PLS model and the RMSEP are reduced relatively. It can be concluded that the DO pre-processing is an effective method to extract the valid s ignal, filter noise and provide a better interpretation of the original spectroscopic data.