基于激光诱导时间分辨荧光的原油识别方法研究

在柴油、汽油、重质燃料油等成品油和原油等溢油油源的区分方面，荧光光谱结合模式识别手段得到了广泛的应用。传统的三维荧光光谱分析方法虽然能够获得溢油样品丰富的成分信息，但难以适应现场应用的要求，目前还停留在实验室检测的阶段。发展适用于现场应用的原油识别方法，对于海洋溢油污染的快速响应与处理意义重大。面向激光雷达的需要，发展了一种基于激光诱导时间分辨荧光手段、结合支持向量机（SVM）模型的原油识别方法，从时间和波长两个不同维度出发，通过对时间窗口和波长范围的选取进行优化，获得了理想的油种识别准确率。实验结果表明通过选取ICCD探测延时为54～74 ns可以将分类正确率从全谱线数据的83.3%提高到88.1%。通过选取波长范围为387.00～608.87 nm的谱线数据，可将疑似油种的分类正确率从全谱线数据的84%提高到100%。激光荧光雷达在实际工作中，受波浪、运载平台晃动等因素的影响，探测延时会出现一定的波动。本文介绍的分类识别方法通过时间和波长两个维度的筛选，更加适用于现场探测数据的识别，并进一步凸显了原油时间分辨荧光光谱特征，为疑似油种分类识别过程中数据量的压缩提供了重要依据。

Discrimination of Crude Oil Samples Using Laser-Induced Time-Resolved Fluorescence Spectroscopy

The Laser-induced fluorescence spectra combined with pattern recognition method has been widely applied in discrimi-nation of different spilled oil ,such as diesel ,gasoline ,and crude oil .However ,traditional three-dimension fluorescence analysis method ,which is not adapted to requirement of field detection ,is limited to laboratory investigatio ns .The development of oil identification method for field detection is significant to quick response and operation of oil spill .In this paper ,a new method based on laser-induced time-resolved fluorescence combined with support vector machine (SVM ) model was introduced to dis-criminate crude oil samples .In this method ,time-resolved spectra data was descended into two dimensions with selecting appro-priate range in time and wavelength domains respectively to form a SVM data base .It is found that the classification accurate rate increased with an appropriate selection .With a selected range from 54 to 74 ns in time domain ,the classification accurate rate has been increased from 83.3% (without selection) to 88.1% .With a selected wavelength range of 387.00~608.87 nm ,the classification accurate rate of suspect oil was improved from 84% (without selection) to 100% .Since the detection delay of fluo-rescence lidar fluctuates due to wave and platform swing ,the identification method with optimizing in both time and wavelength domains could offer a better flexibility for field applications .It is hoped that the developed method could provide some useful ref-erence with data reduction for classification of suspect crude oil in the future development .