基于小波系数的近红外光谱局部建模方法与应用研究

局部建模方法使用与预测样本相似的样本建立模型,可解决光谱响应与浓度之间的非线性问题,扩大模型的适用范围,提高预测准确度。采用小波变换进行数据压缩并利用小波系数之间的欧氏距离作为光谱相似性的判据,实现了近红外光谱定量分析的局部建模方法,避免了样本之间的依赖性。将所建立的方法用于烟草样品中氯含量的测定,100次重复计算得到的预测集均方根误差(RMSEP)平均值为0.0665,标准偏差(σ)为0.0045,优于全局建模和基于主成分的局部建模方法。

Local Regression Method in Wavelet Domain and Its Application in Near-infrared Quantitative Analysis

Local regression method was proposed to deal with the non-linear problem in multivariate analysis so as to improve the universality and prediction precision of the models.A local regression method based on wavelet transform for near-infrared(NIR)spectral quantitative analysis was proposed.In this algorithm,discrete wavelet transform(DWT)was firstly utilized to compress the NIR spectra,and then,the calibration subsets were individually selected for each prediction sample according to the Euclidean distance in wavelet domain.It was found that the Euclidean distance in the wavelet domain can more effectively measure the similarity of the samples than the distance in principal components(PCs)space.An application of the local regression method to quantitative determination of chlorine in tobacco samples was investigated.Results show that,over 100 repeated runs,the mean value of root mean square error of prediction(RMSEP)of the prediction set is 0.0665 and the corresponding standard deviation is 0.0045,which are superior to the results of global PLS method and PCA based local regression method.