地震滑坡信息提取方法研究——以2017年九寨沟地震为例

An information extraction method of earthquake-induced landslide:a case study of the Jiuzhaigou earthquake in 2017

地震滑坡是最为常见且破坏力最强的地震地质次生灾害之一,快速准确地获取地震滑坡的分布和范围信息对震后应急救援和灾后重建工作具有重要意义。采用Sentinel-2A遥感影像,以"8·8九寨沟地震"核心震区——漳扎镇为研究区,通过结合变化检测技术和面向对象方法提取地震滑坡信息,利用改进的区域生长算法对提取的滑坡边界进行优化得到最终提取结果。通过与传统CVA提取结果对比,评价分析本文算法的精度。实验验证结果表明,本文算法的地震滑坡体提取总精度在90%以上,Kappa系数优于0.7,提取结果与目视解译结果较为一致。且自动化程度较高,速度较快,满足地震应急救援的时间和精度需求。

Earthquake-induced landslide is the most common and the most destructive geological secondary disaster triggered by the earthquake.Rapid and accurate extraction of the distribution and extent information of landslides is of great significance for post-earthquake emergency rescue and post-disaster reconstructions.In this study,we take the Zhangzha Town,the heavily hit district by the"8·8 Jiuzhaigou earthquake",as the study region,and the information concerning the distribution and extent of the earthquake-induced landslides was extracted by combining the change detection technology and the object-oriented method.The extracted information was optimized using the improved region growing algorithm based on Sentinel-2 A image.The extraction performance was evaluated by comparison with the traditional CVA extraction results,and the results indicated that the overall accuracy of our extraction was above 90%and the Kappa coefficient was over 0.7.The higher automation level and quicker processing of the proposed method in this study meet the time and precision requirements for earthquake emergency rescues.