STAND: Semantic Anchoring Constraint with Dual-Granularity Disambiguation for Remote Sensing Image Change Captioning

2026-04-25

AI Summary

This work addresses the challenge of inaccurate change description in remote sensing imagery caused by multiple sources of ambiguity, including viewpoint variation, scale inconsistency, and insufficient semantic priors. To resolve these ambiguities in an interpretable manner, the authors propose a disambiguation framework that purifies temporal features through semantic anchoring constraints, integrates global context and frequency-domain refocused attention via a macro–micro dual-granularity mechanism, and leverages linguistic category priors to guide the decoding process. This systematic approach effectively mitigates multi-source ambiguities and achieves state-of-the-art performance on several benchmark datasets for remote sensing change description, significantly enhancing both descriptive accuracy and semantic consistency.

Technology Category

• Computer Vision: Remote Sensing / Geospatial AI
• Knowledge Representation and Reasoning: Action, Change, and Causality
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Remote sensing image change captioning (RSICC) aims to describe the difference between two remote sensing images. While recent methods have explored video modeling, they largely overlook the inherent ambiguities in viewpoint, scale, and prior knowledge, lacking effective constraints on the encoder. In this paper, we present STAND, a Semantic Anchoring Constraint with Dual-Granularity Disambiguation for RSICC, to progressively resolve these ambiguities. Specifically, to establish a reliable feature foundation, we first introduce an interpretable constraint to regularize temporal representations. Operating on these purified features, a dual-granularity disambiguation module resolves spatial uncertainties by coupling macro-level global context aggregation for viewpoint confusion with micro-level frequency-refocused attention for small-object scale enhancement. Ultimately, to translate these visually disambiguated features into precise text, a semantic concept anchoring module leverages language categorical priors to tackle knowledge ambiguity during decoding. Extensive experiments verify the superiority of STAND and its effectiveness in addressing ambiguities.

Problem

• remote sensing image change captioning
• ambiguity
• viewpoint
• scale
• prior knowledge

Innovation

• Semantic Anchoring
• Dual-Granularity Disambiguation
• Remote Sensing Image Change Captioning
• Temporal Representation Regularization
• Frequency-Refocused Attention