Multi-Semantics-Aware Optimal Feature Matching for Multimodal Remote Sensing Images

Read full text Summarise with VIRAT
Authors: Yun Liao, Yide Di, Yubing Liu, Hao Zhou, Xuewen Tan, Junhui Liu, Qing Duan
Cited by: 0
References: 42

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Feature matching for multi-modal remote sensing images is a critical and challenging task due to the substantial differences in imaging mechanisms across diverse sensors. Traditional feature matching methods often struggle to achieve high accuracy under such conditions. To better exploit the structural and semantic information inherent in multi-modal remote sensing data, we propose MSAMatch, a Multi-Semantics-Aware optimal feature Matching method. MSAMatch incorporates multi-level and multi-semantic attention mechanisms to enhance cross-modal feature extraction and information interaction, thereby significantly improving matching accuracy. Furthermore, we introduce a semantic-guided global matching strategy that integrates se mantic cues with local features to enhance robustness. Finally, an optimal fine-level matching is achieved through bidirectional probabilistic modeling based on Gaussian distributions. Extensive experiments on several benchmark multi-modal remote sensing datasets demonstrate that MSAMatch achieves state-of-the-art performance in feature matching tasks. The code will be available at: https://github.com/LiaoYun0x0/MSAMatch

Fields of Study & Subjects

Advanced Image and Video Retrieval Techniques Satellite Image Processing and Photogrammetry Advanced Image Fusion Techniques

Publication details

Journal / source
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Pages
1–18
Publication date
1 January 2026
Language
EN