Identification Of Marine Oil Spill By Fisher Discrimination Model Based On Simultaneous Fluorescence And Wavelet Analysis

Identification of oil spill species has become an important topic for investigation on marine oil spill accidents.Although the identification technology has been widely concerned by scholars at home and abroad,and is mostly qualitative identification using spectral contrast.The accuracy of identification is easily affected by subjective judgments.To solve this problem,the wavelet analysis and Fisher discriminant method were used to establish a discriminant model with high accuracy and pertinence in this study,which can quantitatively identify fuel oil,Middle East crude oil and Non-Middle East crude oil.Eight kinds of marine fuels,seven kinds of the Middle East crude oils and fourteen kinds of Non-Middle East crude oils were collected as the investitation objects,and were detected using synchronous fluorescence spectrometry.The influence of weathering on the fluorescence characteristics of oil was observed by means of the similarity before and after weathering.Oman crude oil was taken as an example for wavelet decomposition according to Daubechies wavelets,Coiflet wavelets,Symmlet wavelets and Bior,wavelets.The correlation coefficients between the reconstructed spectrum and the original spectrum were calculated in terms of different scales and vanishing moments,and two wavelet basis function such as coif4,db7 were finally selected.The fluorescence information at 255±2nm and 354±2nm,which was not detected in original spectrum,was extracted by wavelet analysis,and the fluorescence information at 380±2nm was filtered out since its fluorescence characteristics had no significant and greaty affected by weathering.Two wavelet basis functions of coif4 and db7 were used for Fisher modeling,respectively.The overall accuracy of the F isher discriminant base on coif4 was 89.66%.Among them,8 marine fuels and 7 Middle East crude oils were all correctly identified,and Gabon Rabbi,Xinjiang and Bohai crude oil in 14 Non-Middle Eastern crude oils were mis-identified.The total correct rates of weathered samples through the first and second period were 89.66%.The overall accuracy of the Fisher discriminant base on db7 was 96.6%.Among them,8 marine fuels were all correctly discriminated,all 14 Non-Middle East crude oils were all correctly identified,only Oman crude oil in Middle East crude oils was wrongly identified.The total correct rates of weathering samples through the first and second period were 96.6%.The db7 with higher accuracy was selected to establish the discriminant model The model was verified with 23 kinds of non-modeling oil samples,the accuracy reached 95.7%whether before or after weathering.In this paper,the model establisted by wavelet analysis and Fisher discriminant can not only discriminate crude oil and marine fuel,but also can further subdivid crude oil.The established model will provide a theoretical basis for the real-time and on-line identification of oil spill in the future.