连续小波变换-支持向量回归用于植物样品多组分分析

采用连续小波变换（CWT）技术对近红外光谱（NIR）数据进行预处理，扣除光谱中的背景与噪音成分，再用支持向量回归（SVR）进行建模，建立了用于复杂植物样品多组分分析的建模方法（CWT-SVR），并应用于烟草样品中常规成分（总糖、总植物碱和总氮）含量的测定。结果表明，CWT—SVR方法优于基于全谱数据的SVR和偏最小二乘（PLS）法，为近红外光谱定量分析提供了一种新的建模方法。

Multicomponent analysis of plant samples using continuous wavelet transform and support vector regression

A new approach was proposed for calibration of near-infrared ( NIR) spectroscopy by using support vector regression ( SVR) and continuous wavelet transform (CWT). In the approach, the NIR spectra of plant samples were firstly preprocessed using CWT for denoising and removing the spectral background, then, SVR technique was used for building the calibration model. With application of the method in determination of total sugars, total alkaloids and total nitrogen compounds in tobacco samples, it was shown that the accuracy of the predicted results by the proposed method are better than that by PLS and conventional SVR methods. It maybe an alternative tool for multicomponent determination of complex samples based on NIR spectra.