独立分量分析在近红外光谱定量分析中的应用

采用独立分量分析(ICA)方法,对玉米样品的近红外光谱进行分解,得到统计上独立的各成分光谱;然后用多元回归方法建立基于ICA成分的玉米粗蛋白质、粗淀粉和粗脂肪含量的定量分析模型,3种成分建模集和预测集的化学值和近红外预测值之间的相关系数都较高,且平均相对误差都较低。结果表明,ICA方法建立的玉米样品3个主要成分的近红外模型预测准确度都较高,可应用于玉米育种中大批样品的快速品质分析。

The Application of ICA to Quantitative Analysis by Near Infrared Reflectance Spectroscopy

Independent component analysis was applied to decompose the independent components in near infrared spectral of corn powder samples.Each component spectrum obtained was statistically independent.And quantitative analysis model based on ICA constituents of corn protein,starch and crude fat content was established by means of multiple regressions.The correlation coefficients between the prediction value by NIRS and chemical analysis results of the three components in modeling set and prediction set were relatively high and the mean relative errors were lower.The result shows that the accuracy of the NIR model based on ICA to analysis three main components of the corn samples is high comparatively,and the model can be used in quality analysis in large quantities of maize breeding samples.