基于KOPLS的多组分物质光谱分析方法

KOPLS-based spectral analysis method for multi-component substances

摘要:

为了提升多组分物质浓度的光学测量精度，针对光谱与被测组分的非线性模型，提出了一种基于KOPLS算法的多组分物质光谱分析方法。该方法采用核矩阵将正交无关项转换至高维空间，通过迭代计算与剔除，建立了光谱信号与浓度矩阵之间的非线性回归模型，在保证算法高计算效率的同时解决了传统算法对非线性项分析准确度较低的问题，实现了对多组分物质光谱的高精度分析。通过实验对比了不同算法下的全血样本浓度预测值，实验结果表明KOPLS算法大幅提升了多组分物质浓度计算准确度，实验证明该方法在多组分检测仪器中具有很强的工程应用价值。

Abstract:

In order to improve the optical measurement accuracy, a KOPLS-based spectrum analysis method for multi-component substances is proposed for the nonlinear model between the spectrum and the measured component. This method uses a kernel matrix to convert orthogonal irrelevant items into a high-dimensional space. Through iterative calculation and elimination, a nonlinear regression model between the spectral signal and the concentration matrix is established. It ensures the high computational efficiency of the algorithm and solves the problem of low non-linear regression accuracy in traditional algorithms. This method realizes high-precision spectral analysis of multi-component. Through experiments, the predicted concentration values of whole blood samples are compared under different algorithms. The experimental results show that the KOPLS algorithm greatly improves the prediction accuracy of the concentration of multi-component substances. The experiment proves that this method has strong engineering application value in multi-component measurement instruments.