基于多源遥感数据的汶川地震型滑坡信息提取研究

发布时间：2018-05-07 00:14

本文选题：地震型滑坡 + 高分辨率遥感　；参考：《中国地质大学(北京)》2015年硕士论文

【摘要】：汶川地震引发的大量滑坡给人民生命财产造成巨大损失,引起政府的高度重视和社会的广泛关注。鉴于地震滑坡灾害的严峻形势及其带来的重大损失,积极应对震后滑坡应急调查及灾害评估,对滑坡灾害进行信息提取,快速获取灾害范围及其分布情况,是受灾地区的迫切要求,也是提供应急响应措施及防灾减灾战略重大而迫切的需求,具有十分重要的科学和现实意义。近年来,运用多时相、多分辨率、多传感器的遥感数据及遥感图像处理技术进行灾害信息提取出现在地质灾害防治的各个环节之中,对减灾防灾及灾后重建等方面发挥了不可替代的作用。随着遥感技术的快速发展,高分辨率遥感卫星也取得了长足的进步。高分辨率遥感影像具有丰富的几何纹理信息,能够提供地表大量的细节信息,展开了遥感应用的新纪元。国产高分辨率遥感卫星的相继发射也推动了国产卫星应用的蓬勃发展。本文选取汶川县雁门乡一带作为研究区域,利用资源一号02C、资源三号、高分一号等国产高分辨率遥感影像及地形地貌数据,综合面向对象的信息提取技术开展地震型滑坡信息提取研究。利用尺度估计工具选取相对最优分割尺度,并用多尺度分割技术对影像进行对象分割,以此避免凭经验选取分割尺度的盲目性。在综合分析地震型滑坡自身特点及其在高分辨率影像上成像特征的基础上,利用信息增益法进行特征选择,选取影像光谱特征、纹理特征、几何特征及地形地貌特征等作为分类依据。通过遥感影像上滑坡区域的特征值确定阈值,根据阈值分布特征选取隶属度函数建立提取规则集,对地震型滑坡信息进行提取。提取结果显示,该区地震型滑坡主要沿河流两岸分布且位于断层两侧,在震源周围分布比较密集,通过与人工目视解译结果对比,本文所提出的基于多源数据的面向对象提取方法最大提取精度达72%以上。综合研究表明,运用面向对象的分类方法在多源遥感影像上对地震诱发的突发性滑坡(地表破坏严重,规模大)进行信息提取具有可行性,可为地质灾害及其孕灾环境的宏观调查、灾害体监测与评估以及灾后应急调查提供重要技术支撑,在地震型滑坡信息提取中具有较大的应用潜能。
[Abstract]:A large number of landslides caused by Wenchuan earthquake caused great losses to people's lives and property, which caused great attention by the government and the society. In view of the severe situation of earthquake and landslide disaster and its great loss, the emergency investigation and assessment of landslide after earthquake should be actively dealt with, and the information of landslide disaster should be extracted, and the scope and distribution of landslide disaster should be obtained quickly. It is the urgent demand of the disaster-stricken areas, and it is also the important and urgent need to provide the emergency response measures and the strategy of disaster prevention and mitigation, which has very important scientific and practical significance. In recent years, disaster information extraction using multi-temporal, multi-resolution and multi-sensor remote sensing data and remote sensing image processing technology appears in every link of geological disaster prevention and control. It plays an irreplaceable role in disaster reduction and disaster prevention and reconstruction. With the rapid development of remote sensing technology, high-resolution remote sensing satellite has made great progress. High resolution remote sensing images are rich in geometric texture information and can provide a large amount of detailed information on the surface of the earth, thus opening a new era of remote sensing applications. The successive launches of domestic high-resolution remote sensing satellites also promote the vigorous development of domestic satellite applications. In this paper, Yanmen Township, Wenchuan County, is selected as the research area, and the domestic high-resolution remote sensing images and topographic and geomorphological data, such as Resource-02C, Resource-No.3 and Gaofian No. 1, are used. Based on the object-oriented information extraction technology, the seismic landslide information extraction research is carried out. The scale estimation tool is used to select the relative optimal segmentation scale, and the multi-scale segmentation technique is used to segment the image, so as to avoid the blindness of selecting the segmentation scale based on experience. On the basis of synthetically analyzing the characteristics of seismic landslide and its imaging feature on high resolution image, the feature selection is carried out by using information gain method, and the spectral feature and texture feature of image are selected. Geometric features and landform features are used as the classification basis. The threshold value is determined by the eigenvalue of landslide area in remote sensing image, and the rule set is established by selecting membership function according to the distribution feature of threshold value, and the information of seismic landslide is extracted. The results show that the seismic landslide is mainly distributed along the banks of the river and located on both sides of the fault, and is densely distributed around the epicenter. The results are compared with the results of artificial visual interpretation. The maximum extraction accuracy of the method based on multi-source data is over 72%. The comprehensive research shows that it is feasible to extract the information of earthquake induced sudden landslide (serious ground damage and large scale) by using object oriented classification method in multi-source remote sensing images. It can provide important technical support for macroscopical investigation of geological hazards and their environment, monitoring and evaluation of disaster bodies and post-disaster emergency investigation, and has great application potential in seismic landslide information extraction.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P642.22

【参考文献】