基于三维荧光光谱与图像矩算法的海面溢油检测方法

Detection Method of Spilled Oil at Sea Based on Three-Dimensional Fluorescence Spectrum and Image Moment Algorithm

将三维荧光光谱技术和图像矩算法相结合,提出1种用于海面溢油快速检测的方法。首先,利用4种油类与海水胶束溶液配制24组单一油品及9组混合油品;通过FS920荧光光谱仪采集所有样本的三维荧光光谱(EEM)数据,利用Delaunay插值法去除EEM中的散射光谱;对去散射后的EEM进行灰度化处理,得到对应的EEM灰度图;然后,利用Tchebichef矩、Zernike矩和Krawtchouk矩3种图像矩分别提取EEM灰度图的形状特征;最后,利用聚类分析和逐步回归分别对其特征进行定性定量分析。研究结果表明:基于Tchebichef矩所获得的定性定量结果最佳,单一油品和混合油品最终分类结果均与实际样本设计情况相符合,且相应线性模型相关系数R大于0.99。所提出的方法能够准确对油类样本EEM数据进行解析,达到对其定性定量分析的目的。

A rapid detection method for spilled oil at sea was proposed based on three-dimensional(3D)fluorescence spectra technology and image moments algorithm.First,24 groups of single oil products and 9 groups of mixed oil products were prepared adopting 4 kinds of oil and seawater micelle solution.Second,3D fluorescence spectra(excitation-emission matrix,EEM)of all samples were obtained using FS920 fluorescence spectrometer.The scattering spectra in EEM was removed by Delaunay triangulation method.Meanwhile,grayscale processing was conducted on EEM after removing the scattering spectra to obtain the EEM grayscale image.Next,the shape features of EEM grayscale image were extracted by three image moments,i.e.,Tchebichef moments,Zernike moments and Krawtchouk moments.Finally,qualitative and quantitative analysis of characteristics were performed using cluster analysis and stepwise regression.The research results show that the qualitative and quantitative results obtained based on Tchebichef moments are the best,the final classification results of single and mixed oil products are in accordance with the actual samples design,and the correlation coefficient R of the corresponding linear model is greater than 0.99.The method proposed herein can accurately analyze the EEM data of oil samples and achieve the qualitative and quantitative analysis of oil samples.