相关成分分析法在大米直链淀粉波长选择中的应用

Application of correlative component analysis in the study of selecting wavelength in apparent amylase content with near infrared spectroscopy

为挑选信息含量大、与样品组成或性质相关性较强的光谱区域参与建模,以提高校正模型的精度,采用相关成分分析法对大米直链淀粉的近红外光谱进行分析。结果表明:采用相关成分分析法进行波长选择后,建模波长点数减少为波长选择前的22%,模型预测值与标准值的相关系数R由0.921 2提高到0.973 0,交叉验证标准差(SECV)由3.404 3减小为1.977 4,预测标准差(SEP)由4.810 0减小为1.900 0,模型的预测能力得到显著提高。

Wavelength selecting can be used to select a research space with all combinations of strong correlativity wavelength and large magnitude of the concentration information as final wavelength regions to build a PLS calibration model of NIR. Correlative component analysis algorithm can be employed to identify the magnitude of the information of samples concentration by the variance of the correlative component matrix between spectral matrix and concentration matrix, The apparent amylase content test results showed that the numbers of wavelengths for building the models can be reduced to 22% of the original method, Correlatior~ coefficient can be increased from 0. 921 2 to 0.973 0, standard deviation of cross validation in calibration can be reduced from 3.404 3 to 1. 977 4, and the root mean squared error in prediction was reduced from 4.810 0 to 1.900 0. The prediction precision was greatly improved by correlative component analysis algorithm.