EUVIMEDEUVIMED entdecken
Crossref

Grouped Attention-Based Spectral–Spatial Feature Learning for Collaborative Hyperspectral and LiDAR Classification

Yan Huang, Binghua Lu, Zhizhong Zheng, Kang Ni, Liang Xiao

IEEE Transactions on Geoscience and Remote Sensing · 2026 · 64 · 1-18

DOI: 10.1109/tgrs.2026.3672610 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Geoscience and Remote Sensing, 2026. Verfasst von Yan Huang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 50 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Project funded by China Postdoctoral Science Foundation, Open Research Fund of State Key Laboratory of Information Engineering in Surveying.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The attention mechanism, through its dynamic weight allocation and feature enhancement, effectively improves the representation capability of land-cover features in hyperspectral imaging and light detection and ranging (LiDAR) data. However, most implementations of attention mechanisms tend to neglect the local attributes of land-cover features and the prior information inherent in hyperspectral imagery (HSI) data. Grouped feature learning mechanisms can focus on different characteristics or structures of land-cover objects. Based on this, this article integrates a grouping mechanism into the convolutional neural networks (CNNs)-Transformer framework and proposes a grouped attention-based spectral–spatial feature learning (GAS2-FL) network. Exploiting the low-rank prior of HSI data, the network incorporates a spectral–spatial feature learning module and embeds a grouped feature learning strategy to model spectral information interactions among heterogeneous features. Furthermore, a grouped token mechanism and a nonlocal token learning mechanism are developed to enhance the interaction of local attribute features and the learning of fused representations. The superior performance of the proposed network on four HSI and LiDAR classification benchmarks comprehensively demonstrates that its grouped feature learning mechanism is highly effective at extracting and representing the discriminative features essential for land-cover characterization. The source codes will be available at https://github.com/RSIP-NJUPT/GAS2-FL

Förderung

  • National Natural Science Foundation of China
  • Project funded by China Postdoctoral Science Foundation
  • Open Research Fund of State Key Laboratory of Information Engineering in Surveying
  • Mapping and Remote Sensing, Wuhan University
  • Key Laboratory of Radar Imaging and Microwave Photonics

Zitieren

APA

Huang, Y., Lu, B., Zheng, Z., Ni, K., & Xiao, L. (2026). Grouped Attention-Based Spectral–Spatial Feature Learning for Collaborative Hyperspectral and LiDAR Classification. IEEE Transactions on Geoscience and Remote Sensing, 64, 1-18. https://doi.org/10.1109/tgrs.2026.3672610

Vancouver

Huang Y, Lu B, Zheng Z, Ni K, Xiao L. Grouped Attention-Based Spectral–Spatial Feature Learning for Collaborative Hyperspectral and LiDAR Classification. IEEE Transactions on Geoscience and Remote Sensing. 2026;64:1-18. doi:10.1109/tgrs.2026.3672610

BibTeX
@article{Huang2026,
  author = {Huang, Yan and Lu, Binghua and Zheng, Zhizhong and Ni, Kang and Xiao, Liang},
  title = {Grouped Attention-Based Spectral–Spatial Feature Learning for Collaborative Hyperspectral and LiDAR Classification},
  journal = {IEEE Transactions on Geoscience and Remote Sensing},
  year = {2026},
  volume = {64},
  pages = {1--18},
  doi = {10.1109/tgrs.2026.3672610}
}
Kostenlos starten →