多源光谱信息融合在水质分析中的应用

为解决现有单一光谱法用于水质有机污染综合指标分析精度较低的问题,提出一种基于多源光谱信息融合的水质分析新方法。本方法采用改进的参考独立成分分析方法分别提取紫外吸收光谱和三维荧光光谱对检测水样的多种有效信息特征,去除与水质分析无关的干扰信号;然后采用最小二乘支持向量机进行特征信息融合建模。采用总有机碳指标覆盖范围在3.4~125.3 mg/L内的32个城市地表水和生活污水样本为研究对象,对其紫外光谱、荧光光谱数据进行了信息融合分析实验。结果表明:采用融合分析方法后,对总有机碳指标的分析误差均方根比单一紫外光谱分析法和单一荧光光谱分析法分别下降36.1%和34.7%。

Application of Multi-spectral Information Fusion for Water Quality Analysis

A multi-spectral information fusion based water quality analytical method for the comprehensive index of organic contaminant was proposed to solve the low precision problem of single-spectral analytical techniques.The independent component analysis with reference was improved to extract valid features from the ultraviolet absorption spectrum and the three-dimensional fluorescence spectrum of water samples for interferential signals filtering,and then a least-square support vector machine algorithm was employed to build an information fusion model based on these multi-spectral features.32 samples from surface water and urban wastewater were used to test the proposed UV spectra and fluorescence spectra data fused water analytical method,covering the range of total organic carbon index from 3.4 mg/L to 125.3 mg/L.Compared with these of the UV spectrum analysis method and the fluorescence spectrum analysis method,the root mean square analysis error of the multi-spectral fusion method are decreased by 36.1% and 34.7%,respectively.