多尺度分解和主成分法在近红外光谱分析中的应用

改进近红外光谱分析的传统方法，采取光谱信息二次提取，在利用数学模型进行定量分析之前，先采用小波多尺度分解和主成分分析两种方法对光谱进行两次信息提取，用特征向量表征原光谱，大量缩减建立数学模型的数据量，并建立可靠实用的数学模型。在进行近红外光谱特征提取之前还进行样品聚类分析和光谱去噪处理，避免样品选择和环境因素的影响，为信息提取提供可靠的基础。研究中提出用特征向量表示近红外光谱的方法并在模型应用中验证该向量的有效性。

Use of multi-resolution decomposition and principal components analysis in information abstraction from NIR spectrum

Multi-resolution decomposition based on wavelet transform and principal components analysis are used to abstract the information from NIR spectra before spectra model is built. Eigenvectors of NIR spectra are obtained and the sum of spectra data is greatly reduced. The NIR model based on those eigenvectors is built and the result of prediction shows that the model can be used in practice. In spectra pretreatment, hierarchical cluster analysis and wavelet denoise method are used, so that the risk of sample error and environment disturbance are greatly reduced. The results are verified by spectra model and proved to be very useful.