基于支持张量机回归的三维荧光光谱法水体有机污染物浓度检测  

Detection of dissolved organic matter using three-dimensional fluorescence spectrometry based on support tensor machine regress

作　　者：杜树新[1] 蒋丹红[1] 李林军[1] 

机构地区：[1]浙江大学宁波理工学院信息科学与工程分院,浙江宁波315100

出　　处：《高技术通讯》2014年第1期80-85,共6页Chinese High Technology Letters

基　　金：国家863计划(2009AA04Z123);国家自然科学基金(60974111)资助项目

摘　　要：根据三维荧光光谱二阶张量的数据模式特点,提出了应用基于核函数的支持张量机回归进行三维荧光光谱定量分析的方法,并用其实现了水体有机污染物浓度的检测。在建立回归校正模型中,将二阶张量数据作为模型的输入,充分利用了二阶张量原有的流形结构信息以及数据的内在相关性,提高了模型的推广能力,同时也克服了平行因子法(PARAFAC)、多维偏最小二乘算法等常规二阶校正法需要预先估计组分数、对所预估组分数敏感、要求光谱数据服从三线性分解模型的缺点。对水体中的有机污染物浓度化学耗氧量、总有机碳的检测实验。The kernel function based support tensor machine regression was used to detect dissolved organic matter in wa- ter by using the three-dimensional fluorescence spectrometry. The fluorescence spectrometry with two-order tensor was taken as the input of the calibration model during the model＇ s establishing, and the original manifold structural information and the intrinsic data relationship were fully utilized to increase the calibration model＇ s generalization capability. Moreover, the disadvantages of the traditional methods such as the parallel factor analysis （PARAFAC） and the multi-way partial least squares （N-PLS） in the aspects of need of the component number estimation and sensitivity of the estimated component number and requirements of trilinear decomposition were overcomed. The re- suits of the experiment on detecting total organic carbon （TOC）and chemical oxygen demand （COD）in water showed that the model performance was improved by the proposed ters variation. method and the model was unsensitive to model parame-

关 键 词：光谱学 光谱分析 三维荧光光谱 有机污染物浓度检测 支持张量机 

分 类 号：O433[机械工程—光学工程]