Classification of water contamination developed by 2-D Gabor wavelet analysis and support vector machine based on fluorescence spectroscopy
Classification of water contamination developed by 2-D Gabor wavelet analysis and support vector machine based on fluorescence spectroscopy
复制标题
基于荧光光谱的二维Gabor小波分析和支持向量机的水污染分类
DOI：
10.1364/oe.27.005461
复制
发表时间：
2019-02-18
期刊：
影响因子：
3.8
通讯作者：
Hou, D.
中科院分区：
文献类型：
--
作者：
Huang, P.;Mao, T.;Hou, D.
确定城市供水系统中的特定污染物类别是必要的。传统的检测方法主要是基于常见的水质指标。然而，由于分析时间长、灵敏度不足、需要试剂和产生废液等问题，检测这些水质指标变得困难。这些问题阻碍了高频水的检测和监测。本研究采用三维(3D)荧光光谱作为水质监测方法。提出了一种基于二维Gabor小波和支持向量机多分类的识别方法。采用Delaunay三角插值法对三维荧光光谱进行预处理，从而消除瑞利散射和拉曼散射。提出了一种由不同尺度和旋转角度的滤波器产生的二维Gabor小波函数来提取光谱特征。采用基于Gabor特征描述的块统计方法，提高了描述光谱特征的效率。然后，将多个支持向量机分类器用于污染物分类识别。通过将该方法与常用的特征提取方法主成分分析进行比较，发现二维Gabor小波和分块统计的应用能够有效地描述三维荧光光谱的特征。此外，2D Gabor小波具有很高的分类精度，特别是对于特征峰位置接近或重叠的物质。(C)OSA开放获取出版协议条款下的2019年美国光学学会
The identification of the specific categories of pollutants in the urban water supply system is necessary. Traditional detection methods are based mainly on common water quality indicators. However, inspecting these water quality indicators is made difficult by issues such as long analysis time, insufficient sensitivity, need for reagents, and generation of waste liquid. These problems hinder high-frequency water detection and monitoring. In this study, three-dimensional (3D) fluorescence spectroscopy is adopted as a monitoring method for water quality. An identification method based on two-dimensional (2D) Gabor wavelets and support vector machine (SVM) multi-classification is also proposed. The Delaunay triangulation method for interpolation is used to pre-process 3D fluorescence spectra and thereby eliminate Rayleigh scattering and Raman scattering. A 2D Gabor wavelet function generated by filters of different scales and rotation angles is proposed to extract the features of the spectra. The block statistics method, based on Gabor feature description, is employed to enhance the efficiency in describing spectra features. Then, multiple SVM classifiers are used in pollutant classification and recognition. By comparing the proposed method with principal component analysis, which is a commonly used feature extraction method, this study finds that the application of 2D Gabor wavelets and block statistics can effectively describe the characteristics of 3D fluorescence spectra. Moreover, 2D Gabor wavelets achieve high classification accuracy, especially for substances with closely positioned or overlapping characteristic peaks. (C) 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement