The Identification of the 3-D Fluorescence Spectroscopy Recognition of the Mineral Oil Based on the ICA and the Wavelet Neural Network

摘要

Different kinds of three-dimensional fluorescence spectra of oils have a large degree of overlap. The tradition based on the apparent characteristics of statistical features can only reflect the general characteristics of three-dimensional fluorescence spectra of simple components, or a single fluorescence peak sample differential case is practical, to complex water environment pollution in mineral oil identification has great limitations. The paper use independent component analysis(ICA) algorithm for mineral oil spectral compression dimensionality reduction, feature extraction, extraction of more deep, more elaborate characteristic parameters, and the concentration of information together with the physical significance of the oil pollution of three-dimensional fluorescence spectra characteristic sequence or vector. The mapping relation was obtained by the WNN between the singular value eigenvector and the species of the mineral oil. The WNN realized the recognition of the different kinds of mineral oil. The experiment result indicates that the right of the distinguish rate is 90%.