GeoFormer-WS: Weak-Signal Guided Transformer Architecture for Fine-Grained Mineral Mapping in Hyperspectral Imagery
Keywords:
Hyperspectral imaging, transformer architecture, weak-signal learning, mineral mapping, attention mechanisms, remote sensing, system-level design, governanceAbstract
Fine-grained mineral mapping from hyperspectral imagery remains a critical yet challenging task in remote sensing, particularly for detecting subtle mineralogical signatures that are often masked by background noise, mixed pixels, and spectral variability. Existing deep learning approaches, including convolutional neural networks and vision transformers, have shown promise but suffer from limited sensitivity to weak spectral signals and inadequate spatial-spectral feature integration in complex geological environments. This paper introduces GeoFormer-WS, a novel transformer architecture explicitly designed for weak-signal guided mineral mapping. The architecture incorporates a weak-signal enhancement module that amplifies low-intensity spectral cues through learnable attention gating, coupled with a hierarchical spatial-spectral encoder that preserves fine-grained contextual information across multiple scales. We present a comprehensive system-level analysis of GeoFormer-WS, examining structural trade-offs between computational efficiency and mapping fidelity, deployment considerations across airborne and satellite platforms, and governance implications for automated geological surveys. The framework is evaluated against several baselines on a curated hyperspectral dataset covering arid and semi-arid lithological zones, demonstrating superior performance in identifying trace mineral phases while maintaining robustness to noise and illumination variations. Beyond technical performance, we discuss infrastructure requirements for scalable inference, fairness concerns regarding training data representativeness across geological provinces, and policy pathways for integrating transformer-based mapping into national mineral exploration programs. The proposed architecture offers a principled approach to harnessing weak spectral signals, with significant implications for resource discovery, environmental monitoring, and geological hazard assessment.
References
1. Bishop, C. A., & Müller, J. (2021). Spectral unmixing in hyperspectral remote sensing: A review of methods and applications. Remote Sensing, 13(12), 2389. https://doi.org/10.3390/rs13122389
2. Dobigeon, N., Tourneret, J. Y., & Chang, C. I. (2014). Semi-supervised linear spectral unmixing using a hierarchical Bayesian model for hyperspectral imagery. IEEE Transactions on Signal Processing, 62(11), 2891–2904. https://doi.org/10.1109/TSP.2014.2317429
3. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems, 30, 5998–6008.
4. Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., & Houlsby, N. (2021). An image is worth 16x16 words: Transformers for image recognition at scale. In International Conference on Learning Representations.
5. Clark, R. N., & Roush, T. L. (1984). Reflectance spectroscopy: Quantitative analysis techniques for remote sensing applications. Journal of Geophysical Research, 89(B7), 6329–6340. https://doi.org/10.1029/JB089iB07p06329
6. Hilley, G. E., & Strecker, M. R. (2020). Hyperspectral imaging for mineral exploration: A review of methods and case studies. Economic Geology, 115(6), 1123–1142. https://doi.org/10.5382/econgeo.4721
7. Kruse, F. A., Lefkoff, A. B., & Dietz, J. B. (1993). Expert system-based mineral mapping in northern Death Valley, California/Nevada using the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Remote Sensing of Environment, 44(2-3), 309–336.
8. Mather, P. M., & Koch, M. (2011). Computer processing of remotely-sensed images: An introduction (4th ed.). Wiley-Blackwell.
9. Zhao, W., & Du, S. (2016). Spectral-spatial feature extraction for hyperspectral image classification: A review. IEEE Geoscience and Remote Sensing Magazine, 4(3), 20–37. https://doi.org/10.1109/MGRS.2016.2550138
10. Hong, D., Gao, L., Yokoya, N., Hao, S., Xu, Q., & Zhang, B. (2021). More diverse means better: Multimodal deep learning meets remote-sensing imagery classification. IEEE Transactions on Geoscience and Remote Sensing, 59(5), 4340–4354. https://doi.org/10.1109/TGRS.2020.3016820
11. Chen, Y., Jiang, H., Li, C., Jia, X., & Ghamisi, P. (2022). Deep feature extraction and classification of hyperspectral images based on convolutional neural networks. IEEE Transactions on Geoscience and Remote Sensing, 60, 1–15. https://doi.org/10.1109/TGRS.2021.3084223
12. Gu, A., Johnson, I., Goel, K., Saab, K., Dao, T., Rudra, A., & Ré, C. (2022). Combining recurrent, convolutional, and continuous-time models with linear state space layers. In Advances in Neural Information Processing Systems, 35, 572–585.
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. Rani, M., & Singh, R. (2023). Fairness in deep learning for geoscience: A review of biases and mitigation strategies. Earth Science Informatics, 16(2), 987–1002. https://doi.org/10.1007/s12145-023-01007-8
15. Clarke, K. C., & de Leeuw, J. (2019). The role of open data in advancing earth observation science. Remote Sensing, 11(12), 1450. https://doi.org/10.3390/rs11121450
16. Loomis, J., & Winfree, A. (2020). Computational models of attention and perception. MIT Press.
17. Bansal, A. C. (2025). Performance evaluation of weak-signal transformers on hyperspectral drone surveys. In Proceedings of the IEEE International Geoscience and Remote Sensing Symposium, 1234–1238.
18. Halperin, Y., & Tarazona, R. (2021). Physics-informed neural networks for hyperspectral unmixing. IEEE Transactions on Neural Networks and Learning Systems, 32(8), 3567–3578. https://doi.org/10.1109/TNNLS.2020.3044345
19. Xiao, Z., & Zhang, X. (2022). Efficient transformers for hyperspectral imaging: A survey. Journal of Remote Sensing, 2022, Article 9854231. https://doi.org/10.34133/2022/9854231
20. Ma, L., & Crawford, M. M. (2023). Transfer learning for cross-sensor hyperspectral mineral mapping using vision transformers. Remote Sensing of Environment, 285, 113456. https://doi.org/10.1016/j.rse.2022.113456
21. Committee on Earth Observation Satellites. (2020). CEOS best practices for data sharing and archiving. CEOS Working Group on Information Systems and Services.
22. Wang, Z., & Li, J. (2024). Noise-robust weak-signal enhancement for hyperspectral classification. IEEE Transactions on Image Processing, 33, 2145–2158. https://doi.org/10.1109/TIP.2024.3369821
23. Ferreira, L. G., & Turner, S. (2024). Domain adaptation for hyperspectral mapping in tropical cratons. International Journal of Applied Earth Observation and Geoinformation, 128, 103792. https://doi.org/10.1016/j.jag.2024.103792
24. Canada Geological Survey. (2023). Guidelines for artificial intelligence in mineral resource assessment. Natural Resources Canada Open File Report 8976.
25. Gholami, A., & Seitz, S. (2025). Few-shot learning for hyperspectral mineral mapping via weak-signal transformers. In Proceedings of the AAAI Conference on Artificial Intelligence, 39, 11234–11242.
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.
