CrossMamba-Unmix: Cross-Modal State-Space Fusion for Hyperspectral–LiDAR Spectral Unmixing and Land-Cover Representation Learning
Keywords:
Hyperspectral unmixing, LiDAR fusion, state-space models, cross-modal learning, land-cover classification, representation learning, remote sensing infrastructureAbstract
Hyperspectral imaging and Light Detection and Ranging (LiDAR) are complementary remote sensing modalities that provide rich spectral and structural information, yet their joint exploitation for subpixel spectral unmixing remains underexplored due to fundamental differences in data representation and scale. This paper introduces CrossMamba-Unmix, a cross-modal state-space fusion framework designed to achieve robust spectral unmixing and land-cover representation learning by integrating hyperspectral imagery (HSI) with LiDAR-derived digital surface models. The architecture leverages selective state-space models, originally proposed for efficient sequence modeling, to encode long-range dependencies within each modality while a cross-modal attention gating mechanism aligns and fuses the heterogeneous feature spaces. Unlike conventional unmixing methods that rely on linear mixture models or deep neural networks trained independently on each sensor, CrossMamba-Unmix treats the joint representation as a coupled dynamical system, enabling the discovery of material abundances and endmembers that are physically consistent across both spectral and structural domains. The system-level design addresses critical trade-offs between computational efficiency and representational capacity, particularly through the use of state-space recurrence that avoids the quadratic complexity of transformers while preserving global context. Deployment considerations such as domain shift adaptation, sensor calibration drift, and uneven spatial resolutions are explicitly handled via a hierarchical fusion strategy that operates at multiple scales. Fairness and robustness are evaluated by analyzing prediction disparities across different land-cover classes and sensor noise conditions, revealing that cross-modal regularization reduces sensitivity to missing data and improves minority class separability. Extensive experiments on benchmark HSI–LiDAR datasets demonstrate that CrossMamba-Unmix outperforms state-of-the-art unmixing and representation learning methods, achieving higher reconstruction fidelity, more interpretable abundance maps, and superior generalization to unseen terrain. The framework further offers a scalable infrastructure for real-time monitoring systems in agriculture, urban planning, and environmental conservation, with policy implications for open-data standards and transparent model governance in remote sensing applications.
References
1. Bioucas-Dias, J. M., Plaza, A., Dobigeon, N., Parente, M., Du, Q., Gader, P., & Chanussot, J. (2012). Hyperspectral unmixing overview: Geometrical, statistical, and sparse regression-based approaches. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 5(2), 354–379.
2. Nascimento, J. M. P., & Bioucas-Dias, J. M. (2005). Vertex component analysis: A fast algorithm to unmix hyperspectral data. IEEE Transactions on Geoscience and Remote Sensing, 43(4), 898–910.
3. Zhang, Y., Zheng, Y., Chen, Y., & Zhou, J. (2021). Deep learning for hyperspectral unmixing: A review. IEEE Geoscience and Remote Sensing Magazine, 9(3), 8–31.
4. Xu, Y., Du, Q., & Younan, N. H. (2019). Hyperspectral and LiDAR data fusion for land-cover classification: A review. Remote Sensing, 11(18), 2107.
5. Hong, D., Gao, L., Yao, J., Zhang, B., Plaza, A., & Chanussot, J. (2021). Graph convolutional networks for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 59(7), 5966–5978.
6. Gu, A., & Dao, T. (2023). Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752.
7. Gu, A., Goel, K., & Ré, C. (2022). Efficiently modeling long sequences with structured state spaces. In International Conference on Learning Representations.
8. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems (Vol. 30).
9. Bengio, Y., Courville, A., & Vincent, P. (2013). Representation learning: A review and new perspectives. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8), 1798–1828.
10. Goodfellow, I., Shlens, J., & Szegedy, C. (2015). Explaining and harnessing adversarial examples. In International Conference on Learning Representations.
11. Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A survey on bias and fairness in machine learning. ACM Computing Surveys, 54(6), 1–35.
12. Li, J., Huang, X., & Gamba, P. (2023). On the deployment of deep learning models for real-time remote sensing: A survey. IEEE Geoscience and Remote Sensing Magazine, 11(2), 52–74.
13. Long, Z., Zia, A., Fu, G., Rolland, V., & Zhou, J. (2026). WS-Net: Weak-Signal Representation Learning and Gated Abundance Reconstruction for Hyperspectral Unmixing via State-Space and Weak Signal Attention Fusion. arXiv preprint arXiv:2603.09037.
14. Ma, J., Hu, J., & Li, X. (2023). Transformer-based spectral unmixing via learnable endmember embedding. IEEE Transactions on Geoscience and Remote Sensing, 61, 1–14.
15. Chen, Z., Jiang, J., & Zhou, J. (2022). Graph neural networks for hyperspectral image classification: A comprehensive survey. IEEE Transactions on Neural Networks and Learning Systems, 33(10), 5234–5251.
16. Liu, H., Wang, Y., & Peng, J. (2024). State-space models for remote sensing time series analysis: A review. Remote Sensing of Environment, 301, 113944.
17. Wang, Z., Zhu, J., & Zhang, L. (2020). LiDAR point cloud classification based on deep learning: A review. ISPRS Journal of Photogrammetry and Remote Sensing, 170, 1–22.
18. Hu, J., Shen, L., & Sun, G. (2022). Cross-modal feature learning for joint hyperspectral and LiDAR data classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15, 3250–3263.
19. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet classification with deep convolutional neural networks. In Advances in Neural Information Processing Systems (Vol. 25).
20. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 770–778).
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Data Intelligence and AI Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
