近红外光谱的小波特性研究

通过对在柴油近红外光谱中加入模拟噪声和模拟背景构成的模拟光谱进行小波分析，考察了模拟噪声和模拟背景(线性、二阶和三阶非线性漂移)在小波时-频区域的分布。结果表明，近红外光谱的背景主要分布在低频部分，尤其是逼近部分;噪声主要分布在高频细节部分;反映柴油性质变化的信号主要分布在中间频率细节部分。利用小波变换可同时有效地扣除背景和噪声;选取中间频率细节的小波系数作为多元校正变量进行柴油性质分析，有利于提高分析精度和分析速度。

Wavelet Property Analysis of Near Infrared Spectra

Wavelet analysis(WT) was conducted on the simulated near infrared spectra(NIR) obtained by adding simulated background and simulated noise into diesel NIR.Results show that the background components are mainly located in the low frequency region,while noise in the high frequency region,and useful signal in the middle frequency region.Background and noise components can be simultaneously subtracted from the spectra by WT.The WT coefficients in mid-frequency details can be selected as variables to build the multivariate calibration model,which can improve analytic accuracy and reduce the analysis time.