【作者】 钱育（石羡）；

【导师】 左春；

【作者基本信息】 中国科学院研究生院（软件研究所） ， 计算机软件与理论， 2004， 硕士

【摘要】 为了防止自然灾害和减少自然灾害对财产保险造成的损失，需要根据当前和未来财产保险防灾减损的需要，建立科学的灾情预测模型和财产损失评估模型。综合利用遥感、地理信息系统和全球定位系统集成方法，获取不同模型需要的参数，解决财产保险防灾减损中的关键技术与方法。进行有效的灾害信息收集，并利用这些信息进行灾情预测和损失评估，指导保险公司防灾减损工作的顺利进行。 基于有效指导保险公司防灾减损工作顺利进行的目的，文章总结了灾情预测和财产损失评估的研究现状，分析已有的预测和评估方法，提出了基于DEM(Digital Elevation Model)的灾情预测方法和基于RBF(Radial Basis Function)神经网络的财产损失评估方法。灾情预测方法分为灾情初步预测和灾情修正两个部分，灾情初步预测部分依据高程数据建立DEM模型，综合考虑和分析影响灾情的若干因素，得到淹没的区域、深度和持续时间。灾情修正部分采用神经网络模型，以历史灾情情况为样本进行训练，对前面计算的结果进行修正，从而得到理想的结果，使得预测精度进一步提高；财产损失评估方法以保险标的为评估对象，有效利用收集到的信息，运用RBF神经网络方法建立模型并进行财产损失评估。设计并实现了灾情预测模型和财产保险损失评估模型，集成到财产保险防灾减损原型系统中，并在深圳进行示范应用，取得了良好的效果。更多还原

【Abstract】 In order to prevent natural disasters and reduce the loss of property insurance, it is necessary to set up a scientific disaster prognostication model and a property loss evaluation model according to the need of current and future disaster prevention and loss reduction of property insurance. The technique of disaster prevention and loss reduction of property insurance is an integration of remote sensing, geographic information system and global position system. It collects effective disaster information, according to which we perform disaster prediction and loss evaluation, and instruct insurance companies to do the work of disaster prevention and loss reduction.Our goal is to give effective consultation to insurance companies in their work of disaster prevention and loss reduction. This thesis reviews the current research status of disaster prognostication and property loss evaluation. It analyzes the existing prognostication and evaluation methods, and proposes a disaster prognostication method based on DEM (Digital Elevation Model), and a property loss evaluation method based on RBF (Radial Basis Function) neural network. The disaster prognostication method consists of disaster primary prediction part and disaster correction part. The primary prediction part considers and analyzes various factors which affect the situation of the disaster, to get the flooding area, depth and duration. We use neural network model to implement correction part, train it using the samples of history disaster data, and correct the computing result of the former, then get the ideal result, which improves the prognostication precision. The property loss evaluation method targets insurance item as evaluation object. By using the collected data effectively, it builds a model using the method of RBF neural network, and this model is used to evaluate the property loss. This project designs and implements the disaster prognostication model and the property loss evaluation model, and integrates them into a prototype system of disaster prevention and loss reduction of property insurance. The models were applied in Shenzhen as a test, and showed a satisfactory result.更多还原