测绘通报 ›› 2023, Vol. 0 ›› Issue (3): 33-38.doi: 10.13474/j.cnki.11-2246.2023.0068

• 学术研究 • 上一篇    下一篇

基于InSAR和GRU神经网络的不稳定斜坡地表形变预测

潘建平, 蔡卓言, 赵瑞淇, 付占宝, 袁雨馨   

  1. 重庆交通大学智慧城市学院, 重庆 400074
  • 收稿日期:2022-03-25 发布日期:2023-04-04
  • 作者简介:潘建平(1976-),男,博士,教授,主要从事摄影测量与遥感、地质工程方面的研究。E-mail:panjianping@cqjtu.edu.cn
  • 基金资助:
    贵州省地矿局2019年局管科研项目(黔地矿科合201909);中铁隧道局集团2021年度科技创新计划(隧研合2021-16);中铁隧道局集团2022年度科技创新计划(隧研合2022-14);重庆交通大学研究生科研创新项目(CYS22437)

Forecasting the surface deformation of unstable slopes based on InSAR and GRU neural network

PAN Jianping, CAI Zhuoyan, ZHAO Ruiqi, FU Zhanbao, YUAN Yuxin   

  1. School of Smart City, Chongqing Jiaotong University, Chongqing 400074, China
  • Received:2022-03-25 Published:2023-04-04

摘要: 不稳定斜坡地表形变预测对于滑坡灾害防治和预警具有重要意义。现有监测手段覆盖范围小、成本高,相关预测方法局限于单点预测,对历史数据量要求较高。针对上述问题,本文采用小基线集合成孔径雷达干涉测量(SBAS-InSAR)技术进行不稳定斜坡地表形变监测,设计了一种结合InSAR反演结果和门控循环单元(GRU)神经网络的不稳定斜坡地表形变预测方法。首先使用SBAS-InSAR技术对研究区域进行地表形变监测,然后利用获取到的时序形变反演结果,建立GRU模型进行形变规律学习,最后开展不稳定斜坡地表形变预测。试验结果表明,该方法对不稳定斜坡地表形变的预测平均绝对误差为0.678 mm,平均绝对比例误差为2.7%,相比于传统的支持向量回归(SVR)模型,预测效果提升超过30%,工程应用潜力较大。

关键词: 不稳定斜坡, 形变, 预测, SBAS-InSAR, GRU

Abstract: The prediction of surface deformation of unstable slope is crucial to landslide disaster prevention and early warning. The existing monitoring methods have small coverage and high cost. The relevant prediction methods are limited to single point prediction and have high requirements for the amount of historical data. To solve the above problems, this paper uses Small Base line Subset InSAR (SBAS-InSAR) technology to monitor the surface deformation of unstable slope, and designs a method for predicting the surface deformation of unstable slope based on the inversion results of InSAR and Gated Recurrent Unit (GRU) neural network. Firstly, the SBAS-InSAR technology is used to monitor the surface deformation in the study area, then the obtained time series deformation inversion results are used to establish the GRU model to study the deformation law, and finally carry out the surface deformation prediction of unstable slope.The experimental results show that the mean absolute error of this method for predicting the surface deformation of unstable slope is 0.678 mm,and the mean absolute error percentage is 2.7%.Compared with the traditional support vector regression (SVR) model,the prediction effect is improved by more than 30%, and the engineering application potential is great.

Key words: unstable slope, deformation, forecasting, SBAS-InSAR, GRU

中图分类号: