Bitemporal Remote Sensing Change Detection With State-Space Models

Wang, Lukun ORCID: https://orcid.org/0000-0001-9207-7424, Sun, Qihang ORCID: https://orcid.org/0009-0005-5810-1191, Pei, Jiaming ORCID: https://orcid.org/0000-0003-2774-0511, Khan, Muhammad Attique ORCID: https://orcid.org/0000-0001-5723-3858, Al Dabel, Maryam M. ORCID: https://orcid.org/0000-0003-4371-8939, Al-Otaibi, Yasser D. ORCID: https://orcid.org/0000-0002-1464-8401 and Bashir, Ali Kashif ORCID: https://orcid.org/0000-0003-2601-9327 (2025) Bitemporal Remote Sensing Change Detection With State-Space Models. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 18. pp. 14942-14954. ISSN 1939-1404

Preview

Published Version Available under License Creative Commons Attribution. Download (5MB) | Preview

Abstract

Change detection in very-high-resolution remote sensing images has gained significant attention, particularly with the rise of deep learning techniques such as convolutional neural networks and Transformers. The Mamba structure, successful in computer vision, has been applied to this domain, enhancing computational efficiency. However, much of the research focuses on improving global modeling, neglecting the role of local information crucial for change detection. Moreover, there remains a gap in understanding which structural modifications are more suited for the change detection task. This article investigates the impact of different scanning mechanisms within Mamba, evaluating five mainstream methods to optimize its performance in change detection. We propose local bitemporal change detection mamba (LBCDMamba), a novel architecture based on our proposed local–global selective scan module, which effectively integrates global and local information through a unified scanning strategy. To address the lack of fine-grained details in current models, we propose a multibranch patch attention module, which captures both local and global features by partitioning data into smaller patches. In addition, a bitemporal feature fusion module is proposed to fuse bitemporal features, improving temporal–spatial feature representation. Extensive experiments on three benchmark datasets demonstrate the superior performance of LBCDMamba, outperforming existing popular methods in change detection tasks. This work also provides new insights into optimizing Mamba for change detection, with potential applications across remote sensing and related fields.