An Altitude-Conditioned Spectral–Spatial Transformer for Reliable Cross-Altitude Crop Classification from Unmanned Aerial Vehicle Hyperspectral Imagery

Abstract

Crop maps from unmanned aerial vehicle hyperspectral imagery increasingly guide site-specific field decisions, yet the classifiers behind them are almost always trained and evaluated at a single flight altitude. Because operational height varies with weather, battery and airspace limits, the ground sampling distance changes and map accuracy silently degrades at unseen heights, an error rarely measured before maps are trusted. We quantify this cross-altitude reliability gap and identify which design choices contain. Using a public open-field archive of three vegetable crops (tomato, cabbage and eggplant) imaged at six nominal heights (20–100 m), we build a reproducible protocol that trains at 30 m and evaluates at five withheld heights; because trustworthy labels exist only at 30 m, evaluation uses spectral-angle pseudo-labels together with chance-corrected metrics and an independent manual-validation set. We propose an altitude-conditioned spectral–spatial transformer with resolution augmentation and self-supervised pretraining, and compare it against convolutional, transformer and domain-adversarial baselines over five seeds. The proposed model gives the most reliable transfer, reaching a Cohen's kappa of 0.21 against 0.11–0.14 for the baselines, the best result on every chance-corrected metric and across all five seeds, with the firm evidence resting on the 25–50 m band. This advantage is relative and chance-corrected rather than absolute. Because the archive carries no ground truth, evaluation is referenced to Spectral Angle Mapper (SAM) pseudo-labels, so the species-level model ranking is SAM-consistent and awaits field-plot confirmation. An adversarial invariance component leaves flight height fully decodable, so informing the model of the altitude, not enforcing invariance, drives the gain, and no model exceeds a majority-class predictor in overall accuracy. Altitude-robust monitoring cannot yet be assumed: surveys should fly at consistent, documented heights and report chance-corrected accuracy. We release the protocol, pseudo-labeling pipeline, manual-validation set and all code.

Keywords

• Precision agriculture
• Hyperspectral imaging
• Unmanned aerial vehicle
• Crop classification
• Flight altitude

Ethical Statement

Ethics committee approval was not required for this study because there was no study on animals or humans. The study uses only publicly accessible, aggregated open data and does not process directly identifiable personal data.

Thanks

The author received no funding, grants or other support for this study, and thanks the providers of the public hyperspectral archive. The archive is publicly available from Mendeley Data (https://doi.org/10.17632/7r4r98khyw.2); all code, fixed data splits, trained checkpoints and the independent manual-validation set are provided in an anonymized repository (https://anonymous.4open.science/r/HSI-AltitudeRobust) to preserve double-blind review, with the de-anonymized permanent repository released on acceptance.

References

1. Chen, N., Fang, L., Xia, Y., Xia, S., Liu, H., & Yue, J. (2024). Spectral query spatial: Revisiting the role of center pixel in transformer for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 62, 1–14. https://doi.org/10.1109/TGRS.2024.3361652
2. Dong, L., Geng, J., & Jiang, W. (2024). Spectral–spatial enhancement and causal constraint for hyperspectral image cross-scene classification. IEEE Transactions on Geoscience and Remote Sensing, 62, 1–13. https://doi.org/10.1109/TGRS.2024.3360103
3. Estrada, J. S., Çetin, N., Saçılık, K., Ulu, B., Ulu, B., & Auat Cheein, F. (2026). Assessment of avocado water stress using multispectral imaging and deep learning: Toward scalable, cost-effective irrigation monitoring. Expert Systems with Applications, 326, 132688. https://doi.org/10.1016/j.eswa.2026.132688
4. Ganin, Y., & Lempitsky, V. (2015). Unsupervised domain adaptation by backpropagation. Proceedings of the 32nd International Conference on Machine Learning (pp. 1180–1189). PMLR.
5. Guo, X., Feng, Q., & Guo, F. (2025). CMTNet: A hybrid CNN–transformer network for UAV-based hyperspectral crop classification in precision agriculture. Scientific Reports, 15, 12383. https://doi.org/10.1038/s41598-025-97052-w
6. He, K., Chen, X., Xie, S., Li, Y., Dollár, P., & Girshick, R. (2022). Masked autoencoders are scalable vision learners. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 16000–16009). IEEE.
7. Hong, D., Han, Z., Yao, J., Gao, L., Zhang, B., Plaza, A., & Chanussot, J. (2022). SpectralFormer: Rethinking hyperspectral image classification with transformers. IEEE Transactions on Geoscience and Remote Sensing, 60, 5518615. https://doi.org/10.1109/TGRS.2021.3130716
8. Ibañez, D., Fernández-Beltrán, R., Pla, F., & Yokoya, N. (2022). Masked auto-encoding spectral–spatial transformer for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 60, 1–14. https://doi.org/10.1109/TGRS.2022.3217892

9. Kruse, F. A., Lefkoff, A. B., Boardman, J. W., Heidebrecht, K. B., Shapiro, A. T., Barloon, P. J., & Goetz, A. F. H. (1993). The spectral image processing system (SIPS)—Interactive visualization and analysis of imaging spectrometer data. Remote Sensing of Environment, 44(2–3), 145–163. https://doi.org/10.1016/0034-4257(93)90013-N
10. Li, Y., Guo, S., Jia, S., Yan, Y., Jia, H., & Zhang, W. (2025). Quantifying the effects of UAV flight altitude on the multispectral monitoring accuracy of soil moisture and maize phenotypic parameters. Agronomy, 15(9), 2137. https://doi.org/10.3390/agronomy15092137
11. Liu, X., Zhang, Z., Igathinathane, C., Flores, P., Zhang, M., Li, H., Han, X., Ha, T., Ampatzidis, Y., & Kim, H.-J. (2024). Infield corn kernel detection using image processing, machine learning, and deep learning methodologies under natural lighting. Expert Systems with Applications, 238, 122278. https://doi.org/10.1016/j.eswa.2023.122278
12. Munipalle, V. K., Nelakuditi, U. R., Manohar Kumar, C. V. S. S., & Nidamanuri, R. R. (2024). Ultra-high-resolution hyperspectral imagery datasets for precision agriculture applications. Data in Brief, 55, 110649. https://doi.org/10.1016/j.dib.2024.110649
13. Nidamanuri, R. R., Astor, T., Wachendorf, M., Nautiyal, S., Buerkert, A., Dayananda, S., Bhatt, M., Reji, J., Jha, S., Changalagari, M., Chandra, H., Wijesingha, J., & Sarma, A. (2025). High resolution drone-based hyperspectral imagery for precision agriculture applications (Version 2) [Data set]. Mendeley Data. https://doi.org/10.17632/7r4r98khyw.2
14. Pacal, I. (2024). Enhancing crop productivity and sustainability through disease identification in maize leaves: Exploiting a large dataset with an advanced vision transformer model. Expert Systems with Applications, 238, 122099. https://doi.org/10.1016/j.eswa.2023.122099
15. Qin, B., Feng, S., Zhao, C., Xi, B., Li, W., & Tao, R. (2024). FDGNet: Frequency disentanglement and data geometry for domain generalization in cross-scene hyperspectral image classification. IEEE Transactions on Neural Networks and Learning Systems, 36(6), 10297–10310. https://doi.org/10.1109/TNNLS.2024.3445136
16. Roy, S. K., Krishna, G., Dubey, S. R., & Chaudhuri, B. B. (2020). HybridSN: Exploring 3-D–2-D CNN feature hierarchy for hyperspectral image classification. IEEE Geoscience and Remote Sensing Letters, 17(2), 277–281. https://doi.org/10.1109/LGRS.2019.2918719
17. Ulu, B., Özaktan, H., Çetin, N., Jahanbakhshi, A., Ulu, B., Uzun, S., & Uzun, O. (2025). Smart chickpea seed classification system based on machine vision and deep learning along with a review of challenges, limitations and future trends. Smart Agricultural Technology, 12, 101093. https://doi.org/10.1016/j.atech.2025.101093
18. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
19. Wang, H., Ye, Z., Wang, Y., Liu, X., Zhang, X., Zhao, Y., Li, S., Liu, Z., & Zhang, X. (2023). Improving the crop classification performance by unlabeled remote sensing data. Expert Systems with Applications, 236, 121283. https://doi.org/10.1016/j.eswa.2023.121283
20. Wang, Y., Wen, M., Zhang, H., Sun, J., Yang, Q., Zhang, Z., & Lu, H. (2024). HSIMAE: A unified masked autoencoder with large-scale pre-training for hyperspectral image classification. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17, 14064–14079. https://doi.org/10.1109/JSTARS.2024.3432743
21. Wu, G., Al-qaness, M. A. A., AL-Alimi, D., Dahou, A., Abd Elaziz, M., & Ewees, A. A. (2024). Hyperspectral image classification using graph convolutional network: A comprehensive review. Expert Systems with Applications, 257, 125106. https://doi.org/10.1016/j.eswa.2024.125106
22. Zhang, Y., Li, W., Sun, W., Tao, R., & Du, Q. (2023a). Single-source domain expansion network for cross-scene hyperspectral image classification. IEEE Transactions on Image Processing, 32, 1498–1512. https://doi.org/10.1109/TIP.2023.3243853
23. Zhang, Y., Zhang, M., Li, W., Wang, S., & Tao, R. (2023b). Language-aware domain generalization network for cross-scene hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 61, 1–12. https://doi.org/10.1109/TGRS.2022.3233885