水体有毒有机物三维重叠荧光光谱欠定分离方法研究

Polycyclic aromatic hydrocarbon pollutants in the water，causing strong "three" effects-cancer, malformation and gene mutations, exert threat to human health through the food chain. Three-way fluorescence spectroscopy technique is an effective means for fast and real-time monitoring toxic organic pollutants in water. With the unknown chemical compositions and corresponding concentrations of pollutants，analysis of overlapped fluorescence spectra and identification of pollution components are the key problems. For underdetermined problems of overlapped fluorescence spectrum of toxic organic pollutants in water, the research is carried out and solution is proposed in this research subject. Staring from fluorescence spectra characteristics of target pollutants, employing singular value decomposition to determine the number of components, applying derivative spectroscopy and wavelet transform to sparse fluorescence spectra, adopting energy peak and hyper-plane clustering method to estimate the mixed matrix, utilizing generalized inverse of signal reconstruction algorithm to extract single-component fluorescence spectrum，and Identifying the target pollution components according to the correlation coefficient, separating 3-7 kinds of Polycyclic aromatic hydrocarbon pollutants from two mixtures is realizing. Underdetermined separation model and algorithm of two mixtures samples are established to identify several kinds of toxic organic pollutants in water, providing new methods for monitoring toxic organic pollutants in water.

水体多环芳烃类污染物具有致癌、致畸、致基因突变的“三致”效应，通过食物链威胁人体健康。三维荧光光谱技术是快速、实时分析发荧光有毒有机污染物的有效手段之一。在污染物化学组分和相应浓度未知的情况下，进行光谱解析并识别污染组分是其中的关键科学问题。本项目拟开展水体多环芳烃重叠荧光光谱欠定分离解析方法研究。从水体多环芳烃类污染物荧光光谱特征入手，基于奇异值分解确定组分数，基于导数光谱和小波变换对荧光光谱进行稀疏表征，利用能量峰值聚类和超平面聚类方法估计混合矩阵，采用广义逆方法重构单组分光谱，根据相关系数识别目标污染组分，实现从两个观测的混合样本中精准识别3-7个单组分多环芳烃污染物，建立两个重叠荧光光谱的欠定分离模型及相应算法，实现水体多环芳烃类污染物组分同时快速识别，为水质有毒有机污染物监测提供方法。

多环芳烃类污染物具有致畸致癌致基因突变等危害，通过食物链威胁人体健康。三维荧光光谱技术是快速实时分析水体发荧光有毒有机污染物的有效手段。解析三维荧光光谱并识别污染组分是荧光光谱检测技术应用中的关键问题。本项目开展了少样本多组分情况下水体有毒有机污染物荧光光谱欠定分离问题研究，以多环芳烃污染物为代表，基于傅里叶变换、小波变换研究了重叠荧光光谱信号稀疏特性以及稀疏变换方法，基于非负矩阵分解、模糊聚类、广义逆研究了欠定分离模型中单组分光谱信号重构、提取、识别方法；基于奇异值和平行因子方法研究了含有弱信号的荧光光谱欠定分离问题，实现了利用两个混合样本的观测三维荧光光谱分离多种多环芳烃污染组分，解决了观测信号个数小于其中组分数下的污染物单组分光谱恢复提取问题，为荧光光谱法应用于水体有毒有机污染物监测提供了组分准确识别新方法。