Study on Oil Identification Method Based on Three-Dimensional Fluorescence Spectrum Combined With Two-Dimensional Linear Discriminant Analysis
KONG De-ming1, DONG Rui1, CUI Yao-yao2*, WANG Shu-tao1, SHI Hui-chao3
1. School of Electrical Engineering, Yanshan University, Qinhuangdao 066004, China
2. School of Information Science and Engineering, Yanshan University, Qinhuangdao 066004, China
3. School of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, China
Abstract：Oil pollution seriously threatens the natural environment and human health. Therefore, it is very important to identify and deal with oil pollution. Therefore, three-dimensional fluorescence spectroscopy is generally used to detect the presence of oil contaminants in a certain solution. However, the three-dimensional fluorescence spectrum data of oils have high dimensions, and direct analysis is difficult. Therefore, the data dimensionality reduction method can be used to extract the spectral characteristics of the original oil samples. And the obtained spectral characteristics is used to identify and classify the samples. Based on this, the two-dimensional linear discriminant analysis (2D-LDA) is used to extract the characteristics of the oil samples. The differences in the spectral characteristics of the different samples extracted are studied. The obtained spectral characteristics are used as the input of the K nearest neighbor (KNN) classification to obtain the corresponding. Firstly, four different oils samples (diesel, gasoline, aviation kerosene, lubricating oil) was prepared, and each of the oils has 20 samples. So, 80 oils samples were prepared totally. Secondly, three-dimensional (3D) fluorescence spectrum data of all oil samples are collected by an FS920 spectrometer. Then, the spectral data is pre-processed to remove the scattering and to standardize it. Finally, the 2D-LDA algorithm is used to extract the characteristics of the samples, and the KNN algorithm is used to classify. The results were compared between principal component analysis (PCA) and 2D-LDA. 2D-LDA extracted the emission and excitation characteristics. Both accuracy is 95%. However, the accuracy of combining the classification distances of the emission and excitation spectrum characteristics and re-classifying is 100%. It shows that the two types of spectra are complementary to the three-dimensional fluorescence spectrum, and the combination of emission and excitation spectrum characteristics can better classify the sample. The results show that the classification effect of 2D-LDA characteristics extraction is superior to PCA. It shows that 2D-LDA is better for characteristics extraction of 3D-fluorescence spectrum data. Compared with PCA, 2D-LDA uses the intra-class matrix and the inter-class matrix to maximize the projection vector to extract the characteristics of samples. So, the same type of samples are closer, and the different type of samples are separated as much as possible. Therefore, the 2D-LDA can make it easier to identify data after reducing data dimensionality. Its robustness is good. This study provides a reference to identify oils.
Key words：Three-dimensional fluorescence spectra; Two-dimensional linear discriminant analysis; Principal component analysis; K nearest neighbor
孔德明，董 瑞，崔耀耀，王书涛，史慧超. 基于三维荧光光谱结合二维线性判别分析的油类识别方法的研究[J]. 光谱学与光谱分析, 2021, 41(08): 2505-2510.
KONG De-ming, DONG Rui, CUI Yao-yao, WANG Shu-tao, SHI Hui-chao. Study on Oil Identification Method Based on Three-Dimensional Fluorescence Spectrum Combined With Two-Dimensional Linear Discriminant Analysis. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2021, 41(08): 2505-2510.
[1] Chen H, Liu S, Xu X R, et al. Marine Pollution Bulletin, 2015, 90(1-2): 181.
[2] Ng W, Malone B P, Minasny B. Geoderma, 2017, 289: 150.
[3] Damavandi H G, Gupta A S, Reddy C, et al. Conference on Signals, Systems & Computers. IEEE, 2015.
[4] Cui Y Y, Kong D M, Kong L F, et al. IEEE Access, 2020, 8: 17999.
[5] KONG De-ming, ZHANG Chun-xiang, CUI Yao-yao, et al（孔德明, 张春祥, 崔耀耀, 等）. Acta Optica Sinica（光学学报）, 2018, 38(11): 1130005.
[6] KONG De-ming, ZHANG Chun-xiang, CUI Yao-yao, et al（孔德明, 张春祥, 崔耀耀, 等）. Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2019, 39(10): 3129.
[7] YUAN Yuan-yuan, WANG Shu-tao, KONG De-ming, et al（苑媛媛, 王书涛, 孔德明, 等）. Acta Photonica Sinica（光子学报）, 2017, 46(11): 1130002.
[8] Silva A C, Soares S F, Insausti M, et al. Analytica Chimica Acta, 2016, 938: 53.
[9] Imani M, Ghassemian H. Photogrammetric Engineering & Remote Sensing, 2015, 81(10): 777.
[10] Li Q, You J. Multimedia Tools and Applications, 2019, 78：30397.
[11] Chen Z, Zhou L J, Li X D, et al. Procedia Computer Science, 2020, 166: 523.
[12] Swetapadma A, Yadav A. Computers & Electrical Engineering, 2018, 69: 41.
