基于SAR图像的海洋油膜分类研究

发布时间：2019-02-13 00:01

【摘要】：石油资源是当今社会发展不可缺少的重要化石燃料,被誉为“工业的血液”,伴随着海洋石油资源的不断开发利用,接踵而来的海洋水体污染问题日趋严重。在各种海洋污染中,溢油污染位列榜首,对人们的生产生活造成严重的危害。由于溢油事故多发生在天气条件恶劣的情况,因此合成孔径雷达(SAR)作为一种全天候、全天时、高分辨率微波主动成像传感器,已为海上溢油检测的主要手段。本文以溢油SAR影像为研究对象,探究了利用SAR影像进行海洋油膜识别的全过程,并比较了不同方法的处理效果。本文首先对SAR影像进行了预处理,包括辐射矫正、几何校正、滤波等。在此基础上,分别采用Sobel算子、种子填充和单一阈值法对图像进行分割处理,并简单比较了这三种分割方法的优劣。本文的重点部分就在于对溢油图像进行分类。首先采用一般的监督分类方法对图像进行分类处理,继而采用灰度共生矩阵提取纹理信息,在统计分析的基础上筛选出有用的纹理信息,再辅以纹理信息对图像进行分类,并比较了辅以纹理信息前后监督分类方法的分类精度。结果证明辅以纹理特征的图像分类精度要优于一般的监督分类。之后,本文将面向对象分类方法引入了SAR溢油图像的分类中,分别在ENVI-EX和e Cognition两种软件环境下进行了面向对象的海洋溢油图像分类,最终有效识别出了溢油区域,并系统的比较了两种软件的各自特点。本研究的最终目的就在于通过对不同方法、不同软件的应用和对比,为海洋溢油的识别提供一种行之有效的便捷方法,从而服务于海洋环境保护和可持续发展战略。
[Abstract]:Oil resource is an indispensable fossil fuel in the development of society today. It is called "the blood of industry". With the continuous development and utilization of offshore oil resources, the pollution problem of marine water body is becoming more and more serious. Among all kinds of marine pollution, oil spill pollution is the top one, which causes serious harm to people's production and life. As oil spills often occur in bad weather conditions, synthetic Aperture Radar (SAR), as an all-weather, all-day, high-resolution microwave active imaging sensor, has become the main method for oil spill detection at sea. In this paper, oil spill SAR image is taken as the research object, the whole process of ocean oil film recognition using SAR image is explored, and the processing effect of different methods is compared. In this paper, SAR images are preprocessed, including radiation correction, geometric correction, filtering and so on. On this basis, the Sobel operator, seed filling and single threshold method are used to segment the image, and the advantages and disadvantages of these three segmentation methods are compared. The key part of this paper is to classify the oil spill images. Firstly, the general supervised classification method is used to classify the image, then gray-scale co-occurrence matrix is used to extract the texture information. Based on the statistical analysis, useful texture information is selected, and then the image is classified with the texture information. The classification accuracy of the supervised classification method with texture information was compared. The results show that the accuracy of image classification with texture features is better than that of general supervised classification. After that, the object oriented classification method is introduced into the classification of SAR oil spill image, and the object oriented oil spill image classification is carried out under two kinds of software environment, ENVI-EX and e Cognition, finally, the oil spill area is effectively identified. The characteristics of the two kinds of software are compared systematically. The ultimate purpose of this study is to provide an effective and convenient method for the identification of marine oil spills through the application and comparison of different methods and software to serve the marine environmental protection and sustainable development strategy.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】