海洋灾害给国民经济和沿海居民的生命财产安全造成了巨大的损失与影响。历年关于海洋灾害损失信息的获取，都是当地海洋部门通过实地调查、逐级上报的方式完成，需要耗费大量的财力和时间。随着航天技术和传感器技术的迅速发展，航空相机等所获取的地面影像的分辨率也越来越高，并且可以获得全天候，全方位，实时的观测数据。 因此，采用遥感手段对自然灾害情况进行快速监测成为目前遥感领域的研究热点之一。 本文以近海岸的典型承灾体目标为研究对象，对高分辨率遥感影像中承灾体目标的识别方法进行了研究。首先将本文所利用到的三种特征进行详细的介绍，然后在实现对特征提取的基础上，本文提出了一种多特征加权概率融合的承灾体目标的识别方法：利用支持向量机分别对每一种特征进行分类；利用不同特征输出的支持向量机分类结果来分别计算每个特征的特征权重以及每个样本的分类确定性；最后综合特征权重、分类确定性以及 SVM 概率输出结果，通过最大后验概率进行类别判定，完成对承灾体目标物的提取。 针对线性承灾体目标受灾区段损毁差异性大，没有固定的光谱变化模式，以及由此产生的难以精确识别提取的问题，本文提出一种基于梯度主方向确定性的线性目标损毁区段识别方法。该方法利用单一时相的遥感影像，结合历史矢量信息，在矢量数据的引导下，计算每个检测窗口覆盖下的图像区域内的梯度主方向确定性值，并将梯度主方向确定性值与损毁阈值进行比较，以此判定该区域是否发生损毁。为了测试该方法的可行性，采用了两个不同区域的高分辨率航拍遥感影像对该模型进行验证，试验结果表明该方法简便有效，能够准确地检测出线性目标的损毁区段。

Marine disasters have caused huge losses to the national economy and the safety of the coastal people as well as their property. Over the years, the acquisition of marine disaster information is completed by the local marine department through the way of field survey and report to superior department. However, this method requires a lot of manpower, material resources and time. With the rapid development of aerospace technology and sensor technology, the resolution of remote sensing image acquired by aerial camera is also getting higher, and it can obtain all-weather and real-time observation data. Therefore, the real-time monitoring of natural disasters by means of remote sensing has become one of the hot spots in the field of remote sensing. In this paper, we take affected targets in coastal zone as the research object and study the recognition method of them in the high resolution remote sensing images. First of all, we give a detailed introduction of three features which will used in the following chapters. Based on these three features, we propose a recognition method of typical targets in coastal zone using weighted probability strategy for multi-feature fusion. In the proposed method, SVM (support vector machine) classifier is employed to do the classification for every single feature. Then we calculate feature weight and certainty of SVM classification using the outputs form previous step. The final classification result is determined by the maximum posterior probability based on feature weight, certainty of SVM classification and SVM probability output. On the purpose of extracting accurately identifying results which are caused by linear target in the affected zone exist big differences and there is no fixed pattern of spectral changes, a linear affected target detection method is proposed which is based on the main direction of gradient. A single phase of remote sensing image combined with its historical vector information are used to calculate the uncertainty values of main direction of the gradient for each area under the detection window. The uncertainty values are compared with a threshold value to determine whether damages occurred in the region. In order to test the feasibility of the method, high resolution remote sensing images in two different regions are used to verify the test results. The results show that the method is simple and effective and it can accurately detect the damage zone of qualifying objectives.