摘要
采用独立分量分析方法提取近红外光谱的独立分量和影响矩阵,再用GA—BP神经网络对影响矩阵和浓度矩阵进行建模,提出了基于独立分量一遗传算法.人工神经网络回归的近红外光谱建模方法。分析了独立分量数和网络中间隐层的神经元数对模型性能的影响。采用该方法对小麦样品中的水分、蛋白质、淀粉3种主要成分含量进行测定,水分、蛋白质和淀粉的预测值和参考值之间的相关系数R分别为0.9670、0.9804、0.9674。
A method of model construction based on independent component analysis (ICA) , genetic algorithm (GA), back-propagation artificial neural networks (BP-ANN) regression is proposed for near infrared fast checking. The independent components and the contribution matrix of NIR spectra are extracted by ICA, and then the models of the contribution matrix and concentration matrix are built by GA-BP network. The influence of the numbers of independent components and the neurons in the hidden layer on the properties of the model is analyzed. This method has been applied to the fast determination of the content of three main components in the wheat samples. The correlation coefficients (R) between the reference values and the model predicted values of moisture, protein and gluten contents in validation set are 0. 9670, 0.9804 and 0. 9674, respectively.
出处
《计量学报》
CSCD
北大核心
2010年第3期285-288,共4页
Acta Metrologica Sinica
关键词
计量学
近红外光谱
快速检测
独立分量分析
遗传算法
Metrology
NIR spectroscopy
Fast checking
Independent component analysis
Genetic algorithm