光谱学与光谱分析, 2016, 36 (4): 1100, 网络出版: 2016-12-20
基于光谱反射率数据的水面油种鉴别研究
Identification of Oil Type Using Spectral Reflectance Characteristics
油种 光谱 小波分析 主成分分析 聚类分析 Oil type Spectrum Wavelet analysis Principal component analysis Cluster analysis
摘要
为探讨一种快速、 及时对水上油膜种类进行鉴别的方法, 采用水上油膜反射率光谱数据结合聚类分析方法、 主成分分析方法和小波变换分别对厚度为300, 500和1 000 μm的煤油, 300, 1 000和1 500 μm的润滑油, 50, 300和500 μm的轻柴油和500, 2 000 μm的180#柴油等四种常见溢油油种进行判别研究。 聚类分析结果表明: 采用欧氏距离计算样本间的聚类距离, 在距离L=8.976以上能够将样本正确分类, 准确率100%; 对同一油种油膜而言, 油膜厚度接近的更易聚类; 主成分分析结果表明: 对原始数据、 小波概要系数和小波细节系数分别进行主成分分析, 其中小波细节系数对油种区分效果最佳, 四种油膜样品在主成分得分空间中独立分布。 利用反射率光谱数据结合聚类分析和基于小波细节系数的主成分分析对水上油膜种类的鉴别是可行的。
Abstract
The reflectance spectra of 4 common oil types, kerosene (with the thickness of 300, 500 and 1 000 μm), lubricating oil (with the thickness of 300, 1 000, 1 500 μm), light diesel oil (with the thickness of 50, 300, 500 μm) and 180# diesel ( with the thickness of 500 and 2 000 μm) were analyzed by using cluster analysis and principal component analysis (PCA), in order to explore a fast, timely method for oil type identification. The results of cluster analysis showed that: when the cluster distance between samples was calculated by Euclidean distance and when the distance L=8.976 samples could be correctly classified, the accuracy was up to100%; it also showed the thickness of oil film affected the clustering effects; the principal component analysis showed that: the PCA scores of wavelet detail coefficients had the best result among the original data, the wavelet approximate coefficients and detail coefficients. The methods of using spectral reflectance data combined with cluster analysis and the principal component analysis based on wavelet detail coefficients to identify the type of water film are feasible.
刘丙新, 李颖, 韩亮. 基于光谱反射率数据的水面油种鉴别研究[J]. 光谱学与光谱分析, 2016, 36(4): 1100. LIU Bing-xin, LI Ying, HAN Liang. Identification of Oil Type Using Spectral Reflectance Characteristics[J]. Spectroscopy and Spectral Analysis, 2016, 36(4): 1100.