Oil species identification technique developed by Gabor wavelet analysis and support vector machine based on concentration-synchronous-matrix-fluorescence spectroscopy

Abstract

Concentration-synchronous-matrix-fluorescence (CSMF) spectroscopy was applied to discriminate the oil species by characterizing the concentration dependent fluorescence properties of petroleum related samples. Seven days weathering experiment of 3 crude oil samples from the Bohai Sea platforms of China was carried out under controlled laboratory conditions and showed that weathering had no significant effect on the CSMF spectra. While different feature extraction methods, such as PCA, PLS and Gabor wavelet analysis, were applied to extract discriminative patterns from CSMF spectra, classifications were made via SVM to compare their respective performance of oil species recognition. Ideal correct rates of oil species recognition of 100% for the different types of oil spill samples and 92% for the closely-related source oil samples were achieved by combining Gabor wavelet with SVM, which indicated its advantages to be developed to a rapid, cost-effective, and accurate forensic oil spill identification technique.