无人机高分辨率遥感影像地震滑坡信息提取方法*

(1.西南交通大学 地球科学与环境工程学院,四川 成都 611756; 2.成都市规划信息技术中心,四川 成都 610042; 3.四川大学 水力学与山区河流开发保护国家重点实验室,四川 成都 610065; 4.四川大学 水利水电学院,四川 成都 610065)

信息提取; 深度学习; 迁移学习; 地震滑坡; 无人机影像

Method of earthquake landslide information extraction based on high resolution unmanned aerial vehicle images
FU Xiao1,GUO Jiawei1,LIU Xiuju2,LU Heng3,4,YANG Zhengli3,4,XIANG Xia3,4

(1. Faculty of Geosciences and Environmental Engineering,Southwest Jiaotong University,Chengdu 611756,Sichuan, China)(2. Information Technology Center of Chengdu Planning and Management Bureau,Chengdu 610042,Sichuan,China)(3. State Key Laboratory of Hydraulics and Mountain Rive Engineering,Sichuan University,Chengdu 610065,Sichuan,China)(4. College of Hydraulic and Hydroelectric Engineering,Sichuan University,Chengdu 610065,Sichuan,China)

information extraction; deep learning; transfer learning; earthquake landslide; unmanned aerial vehicle images

备注

选取四川“4·20”芦山地震震区发生的地震滑坡为研究对象,首先构建区域地质特色的无人机影像地震滑坡样本库; 然后引入迁移学习机制完成了无人机高分辨率遥感影像地震滑坡信息提取。研究结果表明地震滑坡提取总体精度达87.2%,能满足地震滑破灾害环境宏观调查,滑坡灾害体监测等应急需求。

In this paper,the earthquake landslide in the earthquake zone of 4.20 Lushan earthquake in Sichuan province was selected as the research object. Firstly,the sample library of earthquake landslide based on Unmanned aerial vehicle(UAV)images was established. Then the transfer learning mechanism was introduced to carry out the earthquake landslide information extraction from UAV images. The results show that the overall accuracy of the landslide extraction method is 87.2%,it can meet earthquake emergency needs such as macroscopic investigation of landslide disaster, monitoring of landslide disaster body.