基于变化检测的北川羌族自治县震后重建监测与动态分析  

Post-earthquake reconstruction monitoring and dynamic analysis of Beichuan Qiang Autonomous County based on change detection

作　　者：雷雨竺 刘瑞[1,2] 曹礼刚[1,2] 杨皓麟 LEI Yuzhu;LIU Rui;CAO Ligang;YANG Haolin(Chengdu University of Technology,Chengdu 610059,China;College of Geophysics,Chengdu University of Technology,Chengdu 610059,China;Xihua University,Chengdu 610039,China)

机构地区：[1]成都理工大学,成都610059 [2]成都理工大学地球物理学院,成都610059 [3]西华大学,成都610039

出　　处：《物探化探计算技术》2025年第2期300-310,共11页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：四川省高等学校人文社会科学重点研究基地项目(RWDL2022-YB001);地质灾害防治与地质环境保护国家重点实验室项目(SKLGP2022K026)。

摘　　要：2008年的汶川地震是迄今为止中国最具破坏性和影响范围最广的地震之一。尽管自地震已经过去15年,北川羌族自治县因其异地重建计划仍受到许多学者的关注。目前,对于异地重建还缺乏定量研究方法。相对于传统的变化检测方法,基于深度学习的方法提供了更高的准确性和客观性。本研究的主要目标是使用搬迁后的北川羌族自治县作为研究区,构建一个多时相的北川县变化检测数据集,并创新性地提出一个基于动态变化的建筑检测框架。首先,在北川县的建筑数据集上评估不同模型的性能,以验证BIT模型的高准确性。然后,预测2008年至2016年期间北川县的重建情况,并可视化预测结果。最后,对2008年至2020年期间北川县的恢复和发展过程进行可视化和定量分析。结果显示,BIT模型在北川县的建筑数据集上实现了最高的准确性,平均交并比(MIoU)为83.40%。此外,统计结果显示2008-2016年灾后重建的建筑物面积为8.44 km^(2)。到达2020年灾后重建的建筑物面积为10.59 km^(2)。研究提出的用于震后城市重建的快速多时相提取方法,可为当地政府进行城市规划提供科学依据。The 2008 Wenchuan earthquake is one of the most destructive and widespread earthquake event in China.Beichuan Qiang Autonomous County continues to attract scholarly attention due to its off-site reconstruction initiatives,even 15 years after the disaster.However,quantitative research approaches are scarce for such reconstruction efforts.Deep learning-based techniques offer enhanced accuracy and objectivity compared to conventional change detection methodologies.This study aims to construct a multi-temporal change detection dataset using the relocated Beichuan Qiang Autonomous County as a case study.Additionally,a framework for dynamic change-based building detection will be proposed.The performance of various models will be evaluated on the building dataset of Beichuan County,confirming the superior accuracy of the BIT model.Then,the reconstruction progress from 2008 to 2016 will be predicted and visualized.Finally,this report presents a visualization and quantitative analysis of the recovery and development of Beichuan County from 2008 to 2020.The results indicate that the BIT model achieved the highest accuracy on the Beichuan County building dataset,with Mean Intersection over Union(MIoU)of 83.40%.Furthermore,the statistical results indicate that the area of reconstructed buildings after disasters increased from 8.44 km^(2) in 2008-2016 to 10.59 km^(2) in 2020.The suggested fast multi-temporal phase extraction method for post-earthquake urban reconstruction can provide a scientific basis for local government to carry out urban planning.

关 键 词：汶川地震 变化检测 震后重建 空间分析 北川羌族自治县 

分 类 号：P315.9[天文地球—地震学]