Vision Mamba in Remote Sensing: A Comprehensive Survey of Techniques, Applications and Outlook (2505.00630v2)

Abstract: Deep learning has profoundly transformed remote sensing, yet prevailing architectures like Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) remain constrained by critical trade-offs: CNNs suffer from limited receptive fields, while ViTs grapple with quadratic computational complexity, hindering their scalability for high-resolution remote sensing data. State Space Models (SSMs), particularly the recently proposed Mamba architecture, have emerged as a paradigm-shifting solution, combining linear computational scaling with global context modeling. This survey presents a comprehensive review of Mamba-based methodologies in remote sensing, systematically analyzing about 120 Mamba-based remote sensing studies to construct a holistic taxonomy of innovations and applications. Our contributions are structured across five dimensions: (i) foundational principles of vision Mamba architectures, (ii) micro-architectural advancements such as adaptive scan strategies and hybrid SSM formulations, (iii) macro-architectural integrations, including CNN-Transformer-Mamba hybrids and frequency-domain adaptations, (iv) rigorous benchmarking against state-of-the-art methods in multiple application tasks, such as object detection, semantic segmentation, change detection, etc. and (v) critical analysis of unresolved challenges with actionable future directions. By bridging the gap between SSM theory and remote sensing practice, this survey establishes Mamba as a transformative framework for remote sensing analysis. To our knowledge, this paper is the first systematic review of Mamba architectures in remote sensing. Our work provides a structured foundation for advancing research in remote sensing systems through SSM-based methods. We curate an open-source repository (https://github.com/BaoBao0926/Awesome-Mamba-in-Remote-Sensing) to foster community-driven advancements.

Vision Mamba in Remote Sensing: A Comprehensive Survey of Techniques, Applications and Outlook

The paper "Vision Mamba in Remote Sensing: A Comprehensive Survey of Techniques, Applications and Outlook" provides an extensive review on the emerging usage of Mamba architectures in the domain of remote sensing. The authors systematically dissect existing studies to offer insights into the adoption and adaptation of Mamba-based techniques, highlighting their foundational principles, application domains, and prospective future directions. Understanding the challenges and opportunities presented by Mamba architectures in remote sensing serves as a catalyst for advancing research in this area.

Core Architectural Concerns

Traditional models, notably Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs), dominate remote sensing but are constrained by notable limitations—CNNs with localized receptive fields and ViTs with quadratic computational complexity. State Space Models (SSMs), particularly the Mamba architecture, introduce a promising architectural paradigm wherein linear computational scaling effectively models global contexts. The paper elaborates on micro-architectural advancements including adaptive scan strategies, hybrid SSM formulations, and macro-architectural integrations. The introduction of CNN-Transformer-Mamba hybrids and frequency-domain adaptations marks a significant shift toward efficient feature extraction and long-range dependency modeling.

Empirical Comparisons

The authors present rigorous benchmarking against conventional CNNs and Transformers across various tasks such as classification, semantic segmentation, and change detection. Notably, Mamba architectures exhibited superior handling of large-scale spatial dependencies and computational efficiency, underscoring their potential as advanced frameworks for remote sensing analysis. The structured taxonomy of innovations and applications illustrates comprehensive advancements in a rapidly evolving landscape.

Challenges and Directions

Despite their usefulness, Mamba architectures confront several obstacles including causality constraints and the need for novel SSM formulations tailored to remote sensing imagery. The paper identifies key challenges as follows:

• Causality: Traditional Mamba operates as a causal system optimized for sequential data, necessitating methods to preserve spatial information loss inherent in remote sensing images.
• SSM Formulations: Innovations in SSM formulation remain nascent, presenting opportunities to develop models specifically suited to remote sensing imagery.
• Multi-modal and Bi-temporal Interactions: Effective multimodal and bi-temporal data interaction using Mamba remains an area for exploration, promising enhanced feature integration and improved task performance.
• 3D Data Processing: Leveraging 3D scan strategies for spectral-rich data, such as hyperspectral images, can foster advancements in spatial-spectral relationship modeling.
• Computational Efficiency: Enhancing computational efficiency via improved hardware-aware algorithms or modifications in SSM formulation offers functional advantages, especially for high-resolution imaging tasks.

Anticipated Developments

Given the computational advantages of Mamba architectures, scaling them for large datasets and diverse applications remains pivotal. The integration of Mamba-based systems in foundation models for remote sensing can substantially improve computational efficiency and generalization capabilities. This prospect aligns with successful LLMs, extending the paradigm to visually guided tasks including image retrieval, VQA, and automated image captioning. The paper underscores the promising role of Mamba architectures in developing future-generation systems, thereby propelling advancements in remote sensing technologies.

Conclusion

Overall, this survey highlights the transformative potential of Mamba architectures in the remote sensing domain. By systematically bridging theoretical SSM constructs with practical applications, the paper establishes a structured foundation for advancing remote sensing capabilities through Mamba-based approaches. Addressing current challenges and exploring innovative directions will further enhance the development of seamless, efficient, and high-performance remote sensing systems. The insights and structured taxonomies provided by the authors enrich the understanding, practices, and developmental strategies for future explorations in this promising field.