联合时序InSAR与Transformer的滑坡形变监测及预测研究
Landslide Deformation Monitoring and Prediction by Integrating Time Series InSAR and Transformer
ES评分 0
| DOI |
10.14075/j.jgg.2025.06.209 |
| 刊名 |
Journal of Geodesy and Geodynamics
|
| 年,卷(期) |
2026, 46(5) |
| 作者 |
|
| 作者单位 |
1 上海市测绘院, 上海, 200063; 2 中国地质大学(北京)土地科学技术学院, 北京, 100083
|
| 摘要 |
现有的滑坡研究多局限于单一形变数据, 忽略了温度、露点温度、降水量和时空相关性等多维环境特征因素对滑坡形变的影响。针对上述问题, 本文提出一种基于时序InSAR的多变量Transformer预测模型。该模型融合多维环境特征与时空相关性数据, 采用小波变换方法将时序形变数据分解为周期项和趋势项, 并构建不同结构的Transformer模型进行预测。结果表明, 该模型预测精度显著优于传统方法, 其RMSE和R2分别达到0.72 mm和0.96。本研究可为滑坡形变预测提供新的思路和方法, 对提升滑坡灾害预警能力具有重要意义。
|
| Abstract |
Existing landslide research is mostly limited to single deformation data, ignoring the influence of multidimensional environmental characteristics factors on landslide deformation, such as temperature, dew point temperature, precipitation, and spatiotemporal correlation. To address this issue, we propose a multivariate Transformer prediction model based on time series InSAR technology. The model integrates multidimensional environmental features and spatiotemporal correlation data, and uses wavelet transform to decompose deformation sequences into periodic and trend components. Different Transformer structures are constructed to predict each component. The results show that the prediction accuracy of this model is significantly better than that of traditional methods, with RMSE of 0.72 mm and R2 of 0.96. This research provides a new approach for landslide deformation prediction and is of great significance for improving early warning capabilities.
|
| 关键词 |
SBAS-InSAR; 深度学习; Transformer网络; 滑坡预测; 多维环境特征
|
| KeyWord |
SBAS-InSAR; deep learning; Transformer network; landslide prediction; multidimensional environmental features
|
| 基金项目 |
|
| 页码 |
628-636 |
张腾军, 杨红磊, 刘友奉.
联合时序InSAR与Transformer的滑坡形变监测及预测研究 [J].
大地测量与地球动力学.
2026; 46; (5).
628 - 636.