光谱预处理方法对硝酸盐测量模型影响实验研究及软件设计

摘要

基于Savitzky-Golay (SG)卷积平滑、正态变量变换(SNV)、一阶导数法、附加散射校正(MSC)和小波去噪(WDS)等信号处理方法，对水中硝酸盐偏最小二乘(PLS)测量模型的影响进行实验研究，采用评估均方差(R MSEE)、预测标准差(R MSEP)、相关系数(R )、预测值与样本浓度值回归关系显著性F检验对预处理效果进行考察。实验结果表明，经SG卷积平滑预处理的PLS模型预测准确性优于其他处理方法。同时编写了该5种光谱预处理方法软件，实现了光谱数据采集与预处理、谱图绘制和光谱保存等功能。%This paper studies the effects of 5 spectral pretreatment methods, including Savitzky-Golay (SG) convolution smoothing, standardized normal variate (SNV), first-order derivative, multiplicative scatter correction (MSC) and wavelet denoising (WDS), on the nitrate partial least squares (PLS) measurement model. The root mean square error of estimation (R MSEE), root mean square error of prediction (R MSEP), correlation coefficient R and regression analysis F-testing for predictive values and actual values are adopted to evaluate the effects of pretreatment. The experiment results show that the prediction accuracy of the PLS model treated by the SG convolution smoothing is higher than that of other spectral methods. Five softwares of spectral pretreatment methods are compiled, achieving the functions of spectrum data acquisition and preprocessing, as well as spectral mapping and preservation.