融合拉曼散射光和荧光信号反演海面溢油厚度的算法研究

随着海洋溢油问题的日益严重，多种遥感技术被用于海面溢油监测，其中激光诱导荧光(LIF)技术是目前被认为最有效的海面溢油探测技术之一。Hoge等基于LIF技术提出了一种利用拉曼散射光评估薄油膜厚度的积分反演算法并广泛应用于海面溢油探测，针对该算法存在误差较大的问题，提出一种融合拉曼散射光和荧光信号评估海面溢油厚度的反演算法。首先利用拉曼散射光信号反演油膜厚度，然后利用该反演结果计算获取溢油油品的荧光特征光谱，最后利用荧光信号反演油膜厚度。文中推导了利用荧光信号反演油膜厚度的算法，给出了油品荧光特征光谱的逼近算法，并给出了利用荧光信号反演油膜厚度的误差分析。通过实验对该方法进行了验证，选用原油和柴油为实验油品，以波长405 nm的激光作为激发光源，采集波长范围为420～700 nm，采集了海水的背景荧光和拉曼散射光信号、实验油品2，5，10和20 μm等不同厚度油膜的光谱信号。将采集数据分为训练集和测试集，利用训练集数据采用梯度下降法获取油品的荧光特征光谱，利用测试集数据分别采用积分拉曼法和该方法反演油膜厚度。采用积分拉曼法，原油不同厚度油膜反演结果的平均误差分别为12.6%，4.6%，4.4%和2.3%，柴油不同厚度油膜反演结果的平均误差分别为14.0%，7.0%，4.2%和3.6%；采用本文方法，原油不同厚度油膜反演结果的平均误差分别为2.5%，2.2%，1.2%和1.1%，柴油不同厚度油膜反演结果的平均误差分别为3.0%，2.4%，2.7%和1.6%。实验结果表明，2 μm油膜反演结果的误差降低最多，原油和柴油2 μm油膜的反演结果误差分别由12.6%和14.0%降低为2.5%和3.0%，其他厚度油膜反演结果的误差也有较大程度的降低，油膜厚度反演结果的误差均小于3%，采用本文算法可以有效提高油膜厚度反演结果的精度。

Algorithm Research on Inversion Thickness of Oil Spill on the Sea Surface Using Raman Scattering and Fluorescence Signal

As the problem of marine oil spills becomes more and more serious,a variety of remote sensing technologies are used to monitor oil spills on the sea surface.Among them,Laser-Induced Fluorescence(LIF)technology is considered one of the most effective oil spill detection technologies.Based on LIF technology,Hoge et al.proposed an integral inversion algorithm based on Raman scattering light to evaluate the thickness of thin oil film,which has been widely used in oil spill detection on the sea surface.Given the large error of the algorithm,an inversion algorithm for evaluating the thickness of oil spills on the sea surface is proposed by using Raman scattering light and fluorescence signals.Firstly,the oil film thickness is inversed by Raman scattering light signal,and then the fluorescence feature spectrum of oil is calculated using the inversion result,and finally,the oil film thickness is inversed by using the fluorescence signal.The algorithm for inversion of oil film thickness using fluorescence signal is deduced,the approximation algorithm of oil fluorescence feature spectrum and the error analysis of oil film thickness inversion using fluorescence signal is given.Experiments verify the method,Crude oil and diesel are selected as the experimental oil and the laser with wavelength of 405 nm is used as the excitation source.The collection wavelength range is 420~700 nm.The background fluorescence and Raman scattering spectra of sea water,the fluorescence spectra of 2,5,10 and 20μm oil films are collected.The collected data are divided into a training set and a test set.The fluorescence feature spectrum of the oil is obtained by gradient descent method using the training set data,and the oil film thickness is retrieved by the Raman integration method and the method in this paper respectively,using the test set.Using Raman integral method,the average errors for different thicknesses of crude oil are 12.6%,4.6%,4.4%and 2.3%,and the average errors for different thicknesses of diesel oil are 14.0%,7.0%,4.2%and 3.6%;Using this method,the average errors for different thickness of crude oil are 2.5%,2.2%,1.2%and 1.1%,and the average errors for different thickness of diesel oil are 3.0%,2.4%,2.7%and 1.6%.The experimental results show that the errors of the 2μm oil film inversion results are reduced the most.The errors of the 2μm oil film inversion results for crude oil and diesel oil are reduced from 12.6%and 14.0%to 2.5%and 3.0%.The errors of the oil film inversion results of other thicknesses are also greatly reduced.The errors of the oil film thickness inversion results are all less than 3%.The algorithm can effectively improve the accuracy of the oil film thickness inversion results.