Scholarlink

<p>Detecting kinematic transitions, often termed trend changes, in Interferometric Synthetic Aperture Radar (InSAR) derived displacement time series is a&#xa0;critical task for geophysical hazard monitoring. This process is frequently impeded by nonlinear deformation patterns, irregular sampling, and significant temporal data gaps. To address this challenge, we introduce an Attention-based Time-Gated Long Short-Term Memory (ATGLSTM) architecture that synergistically integrates temporal gating with a&#xa0;self-attention mechanism to enhance Change Point Detection (CPD), specifically under conditions of data sparsity. The proposed architecture was rigorously trained and evaluated using a&#xa0;combination of realistic synthetic time series, calibrated with European Ground Motion Service (EGMS) data, and real-world observations from the geodynamically complex and data-sparse region of Iceland. Benchmarked against a&#xa0;conventional Convolutional Neural Network (CNN) and a&#xa0;standard Time-Gated LSTM (TGLSTM), the ATGLSTM demonstrated superior performance, achieving an F1-score of 82.40% on simulated data and 72.24% on a&#xa0;manually annotated real-world dataset. These results establish the ATGLSTM as a&#xa0;robust and scalable tool for automated deformation monitoring, holding significant potential for improved hazard assessment and geophysical investigations where SAR observations are discontinuous.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Robust Attention-Based Time-Gated LSTM for Change Point Detection in Challenging InSAR Time Series with Data Gaps

  • 期刊论文

  • 发表日期：

    2026年7月2日

查看全文
  • Seyed Arya Fakhri,
  • Mehran Satari

摘要

Detecting kinematic transitions, often termed trend changes, in Interferometric Synthetic Aperture Radar (InSAR) derived displacement time series is a critical task for geophysical hazard monitoring. This process is frequently impeded by nonlinear deformation patterns, irregular sampling, and significant temporal data gaps. To address this challenge, we introduce an Attention-based Time-Gated Long Short-Term Memory (ATGLSTM) architecture that synergistically integrates temporal gating with a self-attention mechanism to enhance Change Point Detection (CPD), specifically under conditions of data sparsity. The proposed architecture was rigorously trained and evaluated using a combination of realistic synthetic time series, calibrated with European Ground Motion Service (EGMS) data, and real-world observations from the geodynamically complex and data-sparse region of Iceland. Benchmarked against a conventional Convolutional Neural Network (CNN) and a standard Time-Gated LSTM (TGLSTM), the ATGLSTM demonstrated superior performance, achieving an F1-score of 82.40% on simulated data and 72.24% on a manually annotated real-world dataset. These results establish the ATGLSTM as a robust and scalable tool for automated deformation monitoring, holding significant potential for improved hazard assessment and geophysical investigations where SAR observations are discontinuous.