【作者】 李昊；

【导师】 田兆硕；

【作者基本信息】 哈尔滨工业大学 ， 仪器科学与技术， 2021， 硕士

【摘要】 海洋生态系统是地球生态系统的重要组成部分,近年来随着石油资源的开发利用,海洋受石油污染日益严重。石油在自然环境下的降解是一个极其缓慢的过程,需要人为监测溢油情况并干预加速降解过程。遥感监测方法由于其非接触的特点成为溢油监测的重要手段,其中激光诱导荧光遥感监测技术具有主动式探测、可全天候工作、速度快、精度高以及荧光信息独特等特点,已经成为当前最为有效的溢油遥感监测技术之一。首先,本文对激光诱导荧光原理、拉曼散射原理、激光雷达方程以及基于荧光分布的油污识别原理进行了理论研究。根据水拉曼散射特性和油激光诱导荧光特性,在355nm激光的激发下,水拉曼散射光信号会对油荧光信号造成干扰,提出使用以插值算法为核心的荧光基线提取算法和油污识别算法相结合的方法来提高油污种类识别准确率。其次,搭建了远距离水上油污探测系统,主要包括355nm高重频激光器、延时选通部分和光接收部分,实现了对于100m和679.5m处目标水体拉曼散射信号的探测。基于Lab VIEW平台实现远距离水上油污探测系统软件集成,实现系统的人机交互界面,界面包括控制面板、油污识别和显示面板三部分,控制面板由激光器控制、延时器控制和相机控制组成;油污识别部分主要包括标准库调用模块、油膜识别模块和拉曼散射光信号扣除模块;显示面板包括相机原始图像显示、实时光谱显示和油污种类及相应匹配度显示。最后,本文对100m处水面不同厚度柴油、橄榄油、花生油、机油、汽油、玉米油和原油进行探测实验。基于实验采集到的样品光谱建立了355nm激光测油数据库。在数据库相同的条件下,测试不同荧光基线提取算法对识别准确率的影响;在同样进行荧光基线提取的条件下,进行数据库丰富程度对识别准确率影响实验。实验证明,梯度相关系数识别算法在识别不同厚度不同种类覆盖水面油膜时匹配准确率最高,在数据库相同的情况下,三次样条插值算法最适合于进行荧光基线提取,使用荧光基线提取算法进行拉曼散射光扣除前后该算法的识别准确率分别为60.78%和89.22%,荧光基线提取算法将该算法的识别准确率提高了28.44%;在进行拉曼散射光扣除的情况下,将数据库的标准光谱数据的丰富程度提高后,识别准确率提高到98.04%,数据库的丰富程度对于识别准确率有着重要影响。系统成功探测并识别出679.5m出水上机油光谱,为遥感监测并识别溢油奠定了基础。更多还原

【Abstract】 The marine ecosystem is an important part of the earth’s ecosystem.In recent years,with the development and utilization of oil resources,the ocean has been increasingly polluted by oil.The degradation of oil in the natural environment is an extremely slow process,which requires human monitoring of oil spills and intervention to accelerate the degradation process.Remote sensing monitoring methods have become an important means of oil spill monitoring due to their non-contact characteristics.Among them,laser-induced fluorescence remote sensing monitoring technology has the characteristics of active detection,all-weather work,fast speed,high precision,and unique fluorescence information.It has become the most effective at present.One of the technologies of oil spill remote sensing monitoring.First of all,this paper conducts theoretical research on the principle of laser-induced fluorescence,the principle of Raman scattering,the lidar equation,and the principle of oil stain recognition based on fluorescence distribution.According to the water Raman scattering characteristics and the oil laser-induced fluorescence characteristics,under the excitation of 355 nm laser,the water Raman scattering light signal will interfere with the oil fluorescence signal.The fluorescence baseline extraction algorithm and the oil stain recognition algorithm based on the interpolation algorithm are proposed.Combining the methods to improve the accuracy of oil pollution type identification.Secondly,a long-distance oil pollution detection system on the water was built,which mainly includes a 355 nm high-repetition frequency laser,a time-delay gating part and a light receiving part,which realized the detection of the Raman scattering signal of the target water at 100 m and 679.5m.Based on the Lab VIEW platform to realize the software integration of the long-distance oil pollution detection system on the water,and realize the human-computer interaction interface of the system.The interface includes three parts: the control panel,the oil pollution recognition and the display panel.The control panel is composed of laser control,delayer control and camera control;The recognition part mainly includes the standard library calling module,the oil film recognition module and the Raman scattered light signal subtraction module;the display panel includes the camera’s original image display,real-time spectrum display,and oil type and corresponding matching degree display.Finally,this paper conducts detection experiments on diesel,olive oil,peanut oil,motor oil,gasoline,corn oil and crude oil of different thicknesses on the water surface at 100 m.Based on the sample spectra collected in the experiment,a 355 nm laser oil measurement database was established.Under the same conditions of the database,test the effects of different fluorescence baseline extraction algorithms on the recognition accuracy;under the same conditions of fluorescence baseline extraction,conduct experiments on the influence of database abundance on the recognition accuracy.Experiments have proved that the gradient correlation coefficient recognition algorithm has the highest matching accuracy when identifying different thicknesses and different types of oil films covering the water surface.In the case of the same database,the cubic spline interpolation algorithm is most suitable for fluorescence baseline extraction,and the fluorescence baseline extraction algorithm is used for the extraction.The recognition accuracy of the algorithm before and after the subtraction of the Raman scattered light is60.78% and 89.22%,respectively.The fluorescence baseline extraction algorithm improves the recognition accuracy of the algorithm by 28.44%;in the case of subtracting the Raman scattered light,the standard of the database is changed.The spectral data has been increased from 7 to 21,and the recognition accuracy has been increased to 98.04%.The richness of the database has an important influence on the recognition accuracy.The system successfully detected and identified the oil spectrum of 679.5m out of the water,laying a foundation for remote sensing monitoring and identifying oil spills.更多还原