Axial-Relation Guided Fusion State Space Model for Optical-Elevation Sensing Image Segmentation

2026-05-15

AI Summary

This work addresses the limitations of insufficient multi-scale contextual modeling and suboptimal cross-modal fusion in semantic segmentation of multi-source remote sensing imagery by proposing the ARG-Mamba framework. Built upon state space models, ARG-Mamba introduces a multi-scale state space module to jointly capture local details and global dependencies. It further presents, for the first time, an axial relation-guided fusion module that explicitly models the global correlations between optical and elevation modalities along both horizontal and vertical directions. The framework achieves efficient and explicit cross-modal feature fusion while maintaining linear computational complexity. Experimental results demonstrate that ARG-Mamba outperforms current state-of-the-art methods on the ISPRS Vaihingen and Potsdam datasets, achieving high accuracy with favorable computational efficiency.

Abstract

Semantic segmentation of multi-source remote sensing images is a fundamental task for Earth observation applications. Existing methods often struggle with insufficient multi-scale context modeling and suboptimal cross-modal feature fusion, limiting their performance in complex high-resolution scenes. To this end, we propose Axial-Relation Guided Fusion Mamba (ARG-Mamba), a state space model-based framework for optical-elevation remote sensing image segmentation. Specifically, we introduce a Multi-Scale State Space Module to capture both fine-grained local details and global contextual dependencies with linear computational complexity. Moreover, an Axial-Relation Guided Fusion Module is designed to explicitly model global cross-modal correlations along horizontal and vertical axes, enabling efficient feature fusion between optical and elevation modalities. Extensive experiments conducted on the ISPRS Vaihingen and Potsdam datasets demonstrate that our ARG-Mamba consistently outperforms state-of-the-art methods while maintaining favorable computational efficiency. The code will be made publicly available at \url{https://github.com/oucailab/ARG-Mamba}.

Problem

• semantic segmentation
• multi-source remote sensing
• cross-modal fusion
• multi-scale context modeling
• optical-elevation sensing

Innovation

• State Space Model
• Cross-modal Fusion
• Axial Relation
• Remote Sensing Segmentation
• Multi-scale Context