The Discrimination of Fluorescence Spectra of Mineral Oil Based on the FASTICA and SVM

In the process of the use of mineral oil and transportation, it is hard to avoid can have leakage of mineral oil in the water. It will cause a lot of pollution. Oil pollution provenance identification and analysis determination technology research, for water environment protection has important significance. Fluorescence spectrometry has become an effective method for the discrimination of the mineral oil. To distinguish elements complex or fluorspar spectrum have more samples of different overlapping, the EEM matrix fluorspar spectral analysis of mineral oil 3 d singular value feature vector of the information is used for the quality control of mineral oil. The characteristics of the mineral oil spectrum are different. It is difficult to get species identified with a simple formula. In this paper, a multi-wavelet algorithm is used on the noise reduction of the original mineral oil fluorescence spectrum signal. The FASTICA algorithm is used to the feature extraction of mineral oil EEM matrix. The SVM algorithm is implemented to the oil, crude oil, kerosene and diesel four kinds of mineral oil effective classification of fluorescence spectrum identification. The experimental results show that the combination of the two methods, the implementation of the variety of mineral oil well identified. The accuracy of the experimental results recognition is for 97%.