摘要
石油污染的出现,导致生态环境遭到破坏。因此,油类识别方法的研究对于环境的保护具有重要意义。采用荧光光谱法获得石油光谱数据,并对其进行预处理,再通过降维方法来提取特征信息,最后利用模式识别算法进行分类,从而可以实现对油类的定性分析,因此研究一种更高效的数据降维方法以及识别分类算法极其重要。基于三维荧光光谱技术,利用稀疏主成分分析(SPCA)对FS920光谱仪测得的荧光光谱数据进行特征提取,再利用支持向量机(SVM)算法对提取的特征数据进行分类识别,获得了一种更加高效的油类识别方法。首先,利用海水和十二烷基硫酸钠(SDS)配制成浓度为0.1 mol·L-1的胶束溶液,将其作为溶剂配制柴油、航空煤油、汽油以及润滑油各20种不同浓度的溶液;然后,利用FS920光谱仪测得样本溶液的三维荧光光谱数据,对得到的光谱数据进行预处理;最后,对预处理后的数据分别利用SPCA和主成分分析(PCA)进行特征提取,再利用SVM和K最近邻(KNN)两种模式识别算法对特征向量进行分类,最终得到四种模型PCA-KNN, SPCA-KNN, PCA-SVM以及SPCA-SVM的分类结果。研究结果表明,由四种模型得到的分类准确率分别为85%, 90%, 90%和95%,其中,在同种分类算法中,利用SPCA进行特征提取得到的分类准确率均比PCA的准确率高5%,因此可知,SPCA的稀疏性具有突出主要成分的作用,在提取光谱特征时能够减小非必要成分的影响,并且载荷矩阵的稀疏化可以去除变量之间的冗余信息,优化降维特征信息,为后续分类提供更有效的数据特征信息;在同种特征提取算法下,利用SVM算法进行分类得到的分类准确率均比KNN算法得到的准确率高5%,表明SVM算法在分类中更具有优势。因此,本文利用三维荧光光谱技术结合SPCA和SVM算法,实现了对石油的准确识别与分类,为今后对石油污染物的高效检测提供了新思路。
The emergence of oil pollution has destroyed the ecological environment. Therefore, the study of oil identification methods is of great significance to the protection of the environment. Petroleum spectrum data can be obtained by fluorescence spectroscopy. At the same time, the spectrum data is preprocessed, and feature information is extracted by dimensionality reduction. Then the pattern recognition algorithm is used for classification, it can realize the qualitative analysis of oil. However, it is vital to study a more efficient way of data dimensionality reduction and recognition algorithms. Based on the three-dimensional fluorescence spectroscopy technology, this paper uses sparse principal component analysis(SPCA) to extract the features of the fluorescence spectrum data measured by the FS920 spectrometer, and the support vector machine(SVM) algorithm applies for classification and recognition, thereby a more efficient oil identification method is obtained. First, seawater and sodium dodecyl sulfate(SDS) was prepared into a micelle solution with a concentration of 0.1 mol·L-1. It was used as a solvent to prepare solutions of 20 different concentrations of 4 kinds of oil: Diesel oil, Jet fuel, Gasoline and Lubricating oil. Then, the three-dimensional fluorescence spectrum was measured by the FS920 spectrometer, and the data schould be preprocessed. Finally, the pre-processed data is extracted using SPCA, and principal component analysis(PCA), and the feature vectors are classified by SVM and K-nearest neighbor(KNN) two pattern recognition algorithms, the classification results of four models PCA-KNN, SPCA-KNN, PCA-SVM and SPCA-SVM are obtained. The research results show that the classification accuracy rates obtained by the four models are 85%, 90%, 90% and 95% respectively. In the same classification algorithm, the classification accuracy obtained by using SPCA is 5% higher than that of PCA. Therefore, SPCA can better highlight the main components in its sparsity, and the sparsity of the load matrix can re
作者
孔德明
陈红杰
陈晓玉
董瑞
王书涛
KONG De-ming;CHEN Hong-jie;CHEN Xiao-yu;DONG Rui;WANG Shu-tao(School of Electrical Engineering,Yanshan University,Qinhuangdao 066004,China;School of Information Science and Engineering,Yanshan University,Qinhuangdao 066004,China)
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2021年第11期3474-3479,共6页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(61601399,61501394,61771419)
河北省自然科学基金项目(F2016203155,F2017203220)资助。
关键词
三维荧光光谱
特征提取
稀疏主成分分析
支持向量机
Three-dimensional fluorescence spectrum
Feature extraction
Sparse principal component analysis
Support vector machine
