项目名称: 基于拓扑空间表示的我国洪涝灾害时空分布规律挖掘与可视化研究
项目编号: No.41201552
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 地理学
项目作者: 张鹏
作者单位: 民政部国家减灾中心
项目金额: 25万元
中文摘要: 洪涝灾害是我国常年造成损失最为严重的灾种。由于洪涝灾情数据分布的不规则性和复杂性,如何挖掘其隐含的时空分布规律,是当前研究界的研究难点。本项目将立足于民政系统长期积累的我国洪涝灾害灾情数据,基于数学和信息科学领域的最新成果,重点研究基于拓扑表示的洪涝灾损时空分布模式的挖掘与可视化,具体包括:多维度组合下最优灾情数据图像化表示理论;尺度独立且能最优编码灾损信息的抽象图像特征提取方法,以及拓扑空间表示数学模型;拓扑空间中灾损模式的无监督机器学习理论;拓扑度量下最优保持灾损模式分布结构的可视化方法。通过建立模型、理论和算法,本项目将深入挖掘我国洪涝灾害发生频次和灾损模式的多尺度时空规律,评估洪涝灾害与特殊地区的关联关系,为我国洪涝灾害的综合防治提供科学参考依据。进一步,本项目的成果将通过模块化方式在国家自然灾害信息管理系统中进行示范验证,为洪涝灾情信息管理提供智能分析和可视化工具。
中文关键词: 灾害评估;机器学习;洪涝灾害;综合风险评估;
英文摘要: Flood causes the most loss in normal years in China. Due to the irregularity and complexity of flood loss data, how to mine the temporal-spatial patterns hidden in such data has been an open problem in the community of natural hazards. In this grant, based on the rich data collected by civil affairs departments and recent advances in mathematics and information science, we will investigate mining and visualizing the temporal-spatial patterns of flood loss data by using topological representations. Our research includes: optimal image representation model of flood loss data with multi-dimension combination; scale-independent and optimal abstract image feature extraction which can optimally encode the loss data, as well as the mathematical model of topological representation; machine learning theory of loss patterns in topological space; visualization method which can optimally preserve the distribution of loss data under topological measures. By setting up models, theories and algorithms, we will carefully study the mining of occurrence and loss patterns of flood in China. We will also assessment the relationship between flood and voluntary areas. By doing this, we aim to provide technical support to flood disaster relief and reduction in China. Moreover, the research in this grant will be demonstrated and valida
英文关键词: disaster assessment;machine learning;flood hazards;integrated risk assessment;