HiT: History-Injection Transformers for Onboard Continuous Flood Change Detection

2026-01-20

AI Summary

This work addresses the challenge of performing continuous flood change detection on small satellites constrained by limited memory and computational resources. To this end, we propose a lightweight Historical Injection mechanism (HiT), integrated into a Prithvi-tiny-based Transformer architecture, which efficiently incorporates historical context from multi-temporal remote sensing data. The proposed method achieves detection accuracy comparable to that of dual-temporal baselines while reducing image storage requirements by over 99%. Furthermore, it enables real-time inference at 43 frames per second on spaceborne hardware such as the Jetson Orin Nano. This study demonstrates, for the first time, the feasibility of onboard, real-time, and continuous flood monitoring, thereby overcoming a critical bottleneck in disaster response under stringent resource constraints.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Hardware-aware ML
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Natural disaster monitoring through continuous satellite observation requires processing multi-temporal data under strict operational constraints. This paper addresses flood detection, a critical application for hazard management, by developing an onboard change detection system that operates within the memory and computational limits of small satellites. We propose History Injection mechanism for Transformer models (HiT), that maintains historical context from previous observations while reducing data storage by over 99\% of original image size. Moreover, testing on the STTORM-CD flood dataset confirms that the HiT mechanism within the Prithvi-tiny foundation model maintains detection accuracy compared to the bitemporal baseline. The proposed HiT-Prithvi model achieved 43 FPS on Jetson Orin Nano, a representative onboard hardware used in nanosats. This work establishes a practical framework for satellite-based continuous monitoring of natural disasters, supporting real-time hazard assessment without dependency on ground-based processing infrastructure. Architecture as well as model checkpoints is available at https://github.com/zaitra/HiT-change-detection

Problem

• flood detection
• onboard processing
• continuous change detection
• small satellites
• multi-temporal data

Innovation

• History-Injection Transformer
• onboard change detection
• flood monitoring
• satellite edge computing
• temporal context compression