基于ASAR数据的海面溢油信息提取

发布时间：2018-10-20 06:48

【摘要】：由于海上环境复杂,溢油事故发生后,很快发生风化、扩散,如果不能进行及时监测,采取应急措施,其对海洋环境和资源的影响将会非常严重。星载合成孔径雷达(Synthetic Aperture Radar, SAR)进行海面溢油探测,主要是利用海面对微波波段的后向散射强度差别来识别溢油。其具有受天气影响小、探测精度高、覆盖范围广等特点,能够保证对海面溢油目标准确识别的要求。目前星载合成孔径雷达监测溢油的研究,主要面临的问题：一是溢油“假目标”的影响,极大降低了溢油目标的识别精度；二是对于溢油目标空间信息分析不足,缺乏从空间相似性角度考虑溢油目标识别。本文提出的海面溢油信息提取方法,将专家知识同面向对象分类方法相结合,较好地解决了“假目标”的影响；同时,以纹理特征作为分类对象输入,利用面向对象的分类方法,进一步挖掘了溢油目标的二维空间特征。本研究的创新点在于：根据星载SAR海面溢油图像“假目标”的成因、特点及发展趋势建立星载SAR海面溢油“假目标”分类规则,利用分类规则将“假目标”归类,与海面溢油的图像特征和溢油事件发生地的背景信息结合,作为专家知识库,由此实现对假目标的剔除。另一方面,研究认识到现有图像信息提取技术的局限性,通过将“假目标”识别与面向对象分类方法结合,进一步考虑目标的二维空间特征,建立了星载SAR图像海面溢油信息提取监测方案,获得了较好的溢油目标识别效果,这也是本研究的另一创新点。研究以2006年黎巴嫩战争导致的海面溢油事故为例,利用ENVISAT-ASAR数据,对本研究提出的技术方法进行了实验和应用,结果表明：“假目标”专家知识库可以很好地剔除溢油“假目标”,与面向对象的分类方法相结合对星载SAR图像海面溢油信息进行分类,相比不进行“假目标”剔除的方法,大大提高了分类算法的效率和精度。星载SAR卫星作为环境灾害监测的重要工具,得到了世界各国的重视,其发展逐渐以星座化、多波段、多极化为特征,从而为本文提出的星载SAR海面溢油信息提取监测方案提供更好的数据支持,其溢油目标识别精度也会进一步提高。
[Abstract]:Because of the complexity of marine environment, the oil spill accidents occur soon after weathering, diffusion, if not timely monitoring, emergency measures, its impact on the marine environment and resources will be very serious. Spaceborne synthetic Aperture Radar (Synthetic Aperture Radar, SAR) is used to detect oil spills, which is mainly based on the difference of the backscattering intensity of sea surface to microwave wave band. It has the characteristics of small weather influence, high detection precision and wide coverage, which can ensure the accurate identification of oil spill targets on the sea surface. At present, the main problems in the research of Spaceborne synthetic Aperture Radar (SAR) monitoring oil spill are as follows: first, the effect of oil spill "false target" greatly reduces the recognition accuracy of oil spill target; second, the spatial information analysis of oil spill target is insufficient. Lack of spatial similarity to consider oil spill target recognition. In this paper, the method of extracting oil spill information from sea surface combines expert knowledge with object oriented classification method, which solves the problem of "false target", and takes texture feature as the input of classification object. The object-oriented classification method is used to further excavate the two-dimensional spatial features of oil spill targets. The innovation of this study lies in: according to the causes, characteristics and development trend of the "false target" in the spaceborne SAR sea surface oil spill image, the classification rules of "false target" are established, and the "false target" is classified by using the classification rule. Combined with the image feature of the oil spill and the background information of the oil spill event, it can be used as the expert knowledge base to eliminate the false target. On the other hand, in recognition of the limitations of the existing image information extraction techniques, we further consider the two-dimensional spatial features of the target by combining the "false target" recognition with the object-oriented classification method. An oil spill information extraction and monitoring scheme based on spaceborne SAR images was established, and a better target recognition effect was obtained, which is another innovation of this study. Taking the oil spill accident caused by the Lebanon War in 2006 as an example, using the ENVISAT-ASAR data, the technical methods proposed in this study were tested and applied. The results show that the expert knowledge base of "false target" can eliminate the oil spill "false target" well, and combine with the object-oriented classification method to classify the oil spill information on the surface of spaceborne SAR image, compared with the method without "false target" elimination. The efficiency and accuracy of the classification algorithm are greatly improved. Spaceborne SAR satellite, as an important tool of environmental disaster monitoring, has been paid attention to by many countries all over the world. Its development is characterized by constellation, multi-band and multi-polarization. Therefore, it can provide better data support for the oil spill information extraction and monitoring scheme proposed in this paper, and the accuracy of oil spill target recognition will be further improved.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：U698.7

【相似文献】