Remote sensing models select image regions to track changes over time
From Coordinates to Candidate Regions: Temporal Change Localization via Region Selection in Remote Sensing Multimodal LLMs
Computer Vision and Pattern RecognitionComputation and Language
Summary
Identifying exactly where things have changed in satellite images is hard, especially when many small objects are close together. The authors developed a new way for computers to pick out specific areas instead of just giving coordinates, helping the model find changes across multiple images taken over time. Their tests show this method is more reliable for spotting changes while still understanding single pictures well. They also analyze how different parts of the system contribute to its success, making it easier to improve in the future.
What this means in practice
- •For satellite imagery analysts: Improve identification and tracking of changes in satellite images by selecting regions instead of relying on coordinate outputs.
- •For environmental monitoring teams: Better detect and localize environmental changes over time using multispectral satellite image sequences with a more robust region-based method.
Authors
Juwan Chung, Sungjune Park, Yeongyun Kim, Yong Man Ro
Abstract
Remote sensing multimodal large language models (RS-MLLMs) have advanced scene understanding and visual question answering over satellite imagery, yet localizing specific objects or changed regions remains challenging. Existing approaches rely on generating bounding box coordinates as token sequences, which is fragile for the small, densely packed objects common in remote sensing and increasingly error-prone when multiple targets must be localized simultaneously. In this work, we present an RS-specific formulation of the region selection paradigm, previously explored in natural-image MLLMs, and extend it to temporal change localization over multi-image sequences. Our framework employs a text-conditioned region proposal module, encodes each candidate as special tokens carrying per-frame visual features enriched with spatial and temporal cues, and lets the LLM localize targets by selecting region tokens in its response. We construct a multi-task training and evaluation suite spanning localization, referring expression, visual grounding, and understanding tasks across single-image and multi-temporal settings. Experiments show that our approach substantially outperforms coordinate-generation baselines on temporal change localization, while improving single-image visual grounding and maintaining competitive understanding performance. Oracle analysis decomposes the contributions of the region proposer and the LLM selector, providing diagnostic insight unique to this framework. Our code will be available at https://github.com/juwan-kr/RS-RegionSelect.