【作者】 刘志军；

【导师】 张良培； 李平湘；

【作者基本信息】 武汉大学 ， 摄影测量与遥感， 2010， 博士

【摘要】 无居民海岛,是指不属于居民户籍管理的住址登记地的海岛。我国大多数无居民海岛具有独特的区位、丰富的物种、典型的地貌等得天独厚的自然环境；某些海岛还拥有资源丰富的历史遗迹和特有的宗教文化特色等；一些远离大陆的无居民海岛已构成坚固的国防前哨阵地,形成捍卫祖国边疆的海上长城,在维护我国主权和海洋权益方面具有特殊的意义和巨大的作用。长期以来,由于缺乏上位法的支持,无居民海岛开发普遍缺少规划,开发的盲目性很大,在海岛上随意炸山采石、滥采滥伐、倾倒垃圾等,已经造成海岛生态大规模退化和严重破坏。随着《中华人民共和国海岛保护法》的实施,通过对其开展遥感监视、监测,动态了解其开发利用和保护情况,有利于保护海岛生态,维护国家海洋权益。遥感因其可获取大范围数据资料,获取信息的速度快、周期短,受条件限制少、手段多,信息量大等特点,特别是随着遥感数据获取技术的不断发展,遥感影像空间分辨率不断提高。在高分辨率影像上,不仅地物的光谱特征明显,而且其景观的结构、形状、纹理和细节等信息也更加突出。因此,高分辨率遥感影像有利于实时监测无居民海岛的位置、形态、面积、岸线长度和人类活动特征,对于获取海岛的地形、判定疑似海岛的存在和识别具有重要作用,在为管理提供无居民海岛岛体及其周边海域人类活动特征等信息方面具有独特的优势。本文以国家海岛数据库为基础,基于多源多类遥感影像数据,围绕无居民海岛遥感识别、无居民岛体和周边海域人类活动特征遥感识别等无居民海岛遥感监测中的关键问题展开研究。具体研究工作和创新之处体现在：(1)本文从面积大小、所属行政区域和离岸距离等角度,深入分析了我国无居民海岛的特点,并对无居民海岛的岛体和周边海域人类活动特征进行了系统地分析。无居民海岛因其所处的地理位置、面积大小、资源环境条件不同而受到不同程度的人类活动影响,按照其社会服务的功能划分可分为围填海、交通运输、工业用岛、旅游娱乐、仓储、农林牧业开发、公共服务、特殊用途和其他等多种类型。根据对无居民海岛遥感影像的分析,无居民海岛人类活动特征主要有码头、道路、桥梁、实体坝、房屋、人工沙滩、小型水库、人工植被和养殖水面等。(2)针对我国无居民海岛的分布广、面积大小不一等特点,利用多源遥感数据,开展无居民海岛遥感识别研究。主要包括：无居民海岛遥感识别规律研究和无居民海岛岸线提取方法研究。针对无居民海岛面积小,成像环境差等限制条件,研究了无居民海岛空间尺度与识别规律,寻找无居民海岛遥感识别的最优影像分辨率。通过分析发现,无居民海岛遥感普查可采用优于3米分辨率的全色遥感数据进行全覆盖,对重点海岛区域可采用优于1米分辨率的卫星遥感图像或航空遥感影像进行详查。在获取最优分辨率的遥感影像基础上,提出基于Beamlets变换的海岛岸线提取方法,提取海岛岸线,并计算海岛面积。通过实验与Roberts算子、Sobel算子、Laplace算子、Canny算子、基于均值比率的D1算子和D2算子等常用检测方法相比,发现基于Beamlets变换的方法能获得更好的边缘检测结果。在通过边缘检测方法获取的封闭岸线基础上形成海岛面状图元,可以获得相对准确的海岛面积。(3)常规的登岛调查难以快速发现无居民海岛岛体的人类活动情况,但使用高空间分辨率影像能够有效监测无居民海岛的人类活动。面向像元的传统影像数据分类与信息提取方法对于高空分辨率影像的高度细节化特征难以进行有效处理,因此本文在采用Mean shifit多尺度分割算法的基础上,综合模糊分类和最邻近分类器分类方法,构建模糊多分类器模型,使用面向对象的信息提取方法快速获取无居民海岛的岛体人类活动特征。实验结果显示,面向对象的信息提取方法可以充分利用影像的光谱信息、纹理信息和拓补关系,从而可以大大提高分类的精度,减少错误分类的像元。(4)无居民海岛周边海域人类活动,如浮筏式养殖、围垦养殖和围填海工业用海等,在影像上具有明显的纹理特征。为充分利用光学遥感影像完整的纹理信息,本文提出了基于模型的正则化影像融合方法,通过对同一区域的高分辨率遥感影像融合研究试验,结果表明,基于模型的正则化影像融合方法比HSV、Gram schmidt、PCA等遥感影像融合方法进一步增强了特征显示能力,改善了分类精度,能够提高养殖水体的识别精度。(5)针对大区域范围内的无居民海岛人类活动,提出了基于矢量图的集成变化检测方法。通过综合集成多时相的遥感影像和各类矢量数据,采用变化检测的方式,结合面向对象多尺度和模糊分类的方法,可以实现无居民海岛群的人类活动特征信息提取。通过洋山港区综合应用实验发现,由于近年来大规模围填海等人类活动的影响,该区域无居民海岛灭失严重,已极大地影响了大小洋山岛及其周边的生态系统和海洋环境。为解决无居民海岛管理的难题,本文初步建立了无居民海岛及其人类活动特征遥感识别的技术路线,为我国无居民海岛的管理提供了可靠的技术手段,可非常方便快捷地实现无居民海岛人类活动特征的提取。本文有关无居民海岛遥感识别方法研究的成果,已在国家重大专项“海域海岛地名普查”和国家海洋局“国家海岛监视监测系统”项目中得以深入应用,对于我国无居民海岛保护与管理具有重要的价值。更多还原

【Abstract】 The uninhabited islands aren’t the resident registration address. The majority of Chinese uninhabited islands have unique location, rich species, unique landscapes and other unique natural environment. Some of the islands also have rich heritage and unique religious and cultural characteristics, etc. Especially, some of the islands which are far way from mainland could be defending the motherland. For a long time, the lack of support for higher-level laws and the development of uninhabited islands haven’t been planned well. As a result, the ecological environment of some uninhabited islands is extensively degraded and severely damaged. With the promulgation and implementation of Island Protection Law, the island ecosystem, national maritime rights and interests could be protected if the development, utilization and protection of uninhabited island could be monitored dynamically by remote sensing.Remote sensing has many advantages, including access to a large range of data, faster access to information, especially the spatial resolution of remote sensing images improved quickly. In the high-resolution images, not only more obvious features of the spectrum, and its landscape structure, shape, texture and details and other information are also very prominent. Therefore, the high-resolution images could be used to monitor location, shape, area, shoreline length and characteristics of human activities of uninhabited islands.The study to the identification of uninhabited islands and their human activity features is done in this thesis based on national islands database and multi-source remote sensing image data.The major works and contributions of this thesis are as follows:The characteristics of uninhabited islands are deeply analyzed from the size, respective administrative areas and offshore distance of the uninhabited islands. The type of the human activities of the body and their surrounding waters of the uninhabited islands were systematically analyzed according to its geographical location, area, resources, environmental conditions and the adjacent inhabited islands. The uninhabited islands could be classed into traffic, industrial, tourism and entertainment, storage, forestry, animal husbandry, public services, special purposes, and other types. The human activities characteristics of uninhabited islands mainly include dock, roads, bridges, solid dam, housing, artificial beach, small reservoirs, artificial vegetations and water surfaces for breeding, etc.·The study to the identification of uninhabited islands mainly includes the RS identification rules and the coastline extraction of uninhabited islands. Some classical edge detection operators, such as Roberts operator, Sobel operator, Laplace operator, Canny operator, D1 operator, D2 operator and the beamlets transform method are analyzed and compared to extract the coastline. Finally, the area calculation of the uninhabited islands in the 2000 National Geodetic coordinates is discussed. The result shows that 3-meter resolution panchromatic remote sensing data may be used to cover all key regions and 1 meter resolution satellite remote sensing images may be used for the special region. The beamlets transform method and canny operator is more practical to the ectract to the artificial coastline.·This thesis discusses the object-oriented information extraction methods in order to identify the human activity features of uninhabited islands quickly and accurately. Based on the study to the Mean shift and FNES object-oriented multi-scale image segmentation algorithm, the nearest neighbor classifier and fuzzy multi-classifier model are compared to class the type of human activity features of uninhabited islands. The result shows the object-oriented information extraction method can take full advantage of images of spectral, shape and texture information, which could greatly improve the classification accuracy.·The human activities of the surrounding waters of Uninhabited islands could be divided into floating raft culture, land reclamation and reclamation of farming the sea, etc.. In order to take full advantage of the complete texture information of optical remote sensing image, this thesis presents a model-based image fusion method which could be used to the high resolution remote sensing images. The method is compared with HSV, Brovey, Gram schmidt, PCA by using the same region. The result shows the fusion of remote sensing image can be further enhanced display capabilities, improve the classification accuracy. It plays an important role for the calculation of the culture area.·The change comparison method based on the change detection method based on the integration of GIS and RS is proposed in this thesis. Based on the method, the human activities information of uninhabited islands could be extracted through combined with object-oriented multi-scale and fuzzy classification methods in the support of comprehensive integration of remote sensing images and multi-temporal vector data. The comprehensive application test shows that the loss of uninhabited islands in the region is very serious because the affect of the reclamation and other human activities in recent years.·In order to solve the problems of the island management, this thesis has creatively established a technology system to recognize the uninhabited islands and their characteristics of human activity by remote sensing. It could provide reliable technical means for the management of uninhabited islands in China and extract the human activity feature information quickly and easily. The research results have been used in the survey on Marine Geographical Names and the State Island Surveillance Monitoring System project.更多还原