Parameterization of three-dimensional fluorescence spectra based on singular values and oils clustering analysis

Depending on oil components, three-dimensional (3-D) fluorescence spectra of oils can be seen as their "fingerprints". Various kinds of oils can be classified by clustering their feature vectors. Statistic parameters such as the average, standard error, centroid, kurtosis, as well as geometrical distribution can be selected as the feature parameters of 3-D fluorescence spectra of oils forming the feature vector. In this paper, oil clustering based on parameterization using singular values of Excitation-Emission Matrix (EEM) of 3-D fluorescence spectra is studied as a progressive work for identification of complex contaminating oils in water. Clustering analysis of oil is completed with singular values forming the feature vector. As one of convincing results obtained by clustering analysis using Matlab tool, single values of the EEM have an advantage over those statistic parameters of apparent features of 3-D fluorescence spectra in classifying oils.