基于激光诱导荧光光谱的油种识别方法研究

发布时间：2018-10-23 10:28

【摘要】：海上溢油污染是当今全球海洋污染最严重的问题之一,造成了巨大的国民财产损失和环境损害,对海洋环境中的溢油污染进行及时、准确的探测,可以缩小灾害范围、厘清责任归属。由于溢油污染物中包含的多环芳烃及其化合物有着较强的荧光活性,采用激光诱导荧光光谱技术可以对溢油种类进行识别。本文发展了一套基于激光诱导荧光技术结合模式识别手段,识别常见的溢油污染物种类的方法。论文首先简要介绍了背景和选题意义,针对激光荧光雷达设备的国内外发展情况、激光诱导荧光手段在油种检测识别的研究进展进行了调研,并对本文的主要工作内容进行了介绍。此外,介绍了激光诱导荧光光谱技术及论文中用到的算法的基本原理。本文的工作内容包括以下三个方面：第一,基于激光诱导发射光谱的油种识别研究,采集了柴油、汽油、重质燃料油和五种原油的发射光谱,分别用PLS-DA、PCA结合BP-ANN和SVM三种算法识别,比较分类结果和模型训练难度选择SVM模型为后续工作的采用的算法。第二,时间分辨荧光光谱数据的特征范围提取,将测得的时间分辨光谱分别从时间和波长两个方向降维,结合SVM模型对时间窗口和波长范围进行优化,从数据处理的角度凸显原油时间分辨荧光光谱的特性范围。第三,基于时间分辨荧光光谱的油种识别研究,其中特征提取方法为PCA和统计参量提取,识别方法为SVM模型。此外,对以上工作中采用的实验装置、实验样品和光谱采集工作也进行了详细介绍。
[Abstract]:Marine oil spill pollution is one of the most serious problems of marine pollution in the world today, which has caused huge national property loss and environmental damage. Timely and accurate detection of oil spill pollution in the marine environment can narrow down the scope of disasters. Clarify the attribution of responsibility. Because the polycyclic aromatic hydrocarbons (PAHs) and their compounds contained in oil spill pollutants have strong fluorescence activity, laser induced fluorescence spectroscopy can be used to identify the oil spills. In this paper, a method based on laser induced fluorescence (LIF) combined with pattern recognition is developed to identify common types of oil spill pollutants. Firstly, the background and the significance of selecting the topic are briefly introduced. According to the development of lidar equipment at home and abroad, the research progress of laser induced fluorescence in oil detection and recognition is investigated. The main work of this paper is introduced. In addition, the laser induced fluorescence spectroscopy and the basic principle of the algorithm used in this paper are introduced. The work of this paper includes the following three aspects: first, the emission spectra of diesel oil, gasoline, heavy fuel oil and five kinds of crude oil are collected based on the laser induced emission spectrum. PLS-DA,PCA combined with BP-ANN and SVM are used to identify the classification results and the difficulty of model training. The SVM model is chosen as the algorithm for the following work. Secondly, the characteristic range of time-resolved fluorescence spectrum data is extracted, the measured time-resolved spectrum is reduced from time and wavelength direction, and the time window and wavelength range are optimized with SVM model. The characteristic range of time resolved fluorescence spectrum of crude oil is highlighted from the point of view of data processing. Thirdly, the oil species recognition based on time-resolved fluorescence spectrum is studied, in which the feature extraction method is PCA and statistical parameter extraction, and the recognition method is SVM model. In addition, the experimental device, experimental sample and spectrum collection are also introduced in detail.
【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：X55;X834

【参考文献】