Leveraging Satellite Image Time Series for Accurate Extreme Event Detection

2025-02-28

2025 IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW)

AI Summary

To address the challenge in extreme weather event detection where bitemporal methods struggle to distinguish disaster signals from background changes, this paper proposes SITS-Extreme—the first satellite image time-series framework specifically designed for extreme event detection. It integrates pre-disaster multi-stage observations and incorporates deep temporal modeling, multi-temporal feature alignment, and a disaster-sensitivity attention mechanism. Evaluation adopts a hybrid paradigm combining synthetic and real-world data. Compared to strong bitemporal baselines, SITS-Extreme achieves over 23% higher detection accuracy on both synthetic and real datasets, significantly improving cross-hazard generalizability and large-scale scalability. Ablation studies confirm the robust contributions of incremental temporal modeling and each architectural component.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Remote Sensing / Geospatial AI
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Climate change is leading to an increase in extreme weather events, causing significant environmental damage and loss of life. Early detection of such events is essential for improving disaster response. In this work, we propose SITS-Extreme, a novel framework that leverages satellite image time series to detect extreme events by incorporating multiple pre-disaster observations. This approach effectively filters out irrelevant changes while isolating disaster-relevant signals, enabling more accurate detection. Extensive experiments on both real-world and synthetic datasets validate the effectiveness of SITS-Extreme, demonstrating substantial improvements over widely used strong bi-temporal baselines. Additionally, we examine the impact of incorporating more timesteps, analyze the contribution of key components in our framework, and evaluate its performance across different disaster types, offering valuable insights into its scalability and applicability for large-scale disaster monitoring.

Problem

• Detecting extreme weather events using satellite time series
• Improving accuracy by filtering irrelevant environmental changes
• Evaluating framework scalability across diverse disaster types

Innovation

• Leverages satellite image time series
• Filters irrelevant changes effectively
• Improves accuracy over bi-temporal baselines