EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

DA-GSGTNet: Dynamic Aggregation Gated Stratified Graph Transformer for Multispectral LiDAR Point Cloud Segmentation

Qiong Ding, Runyuan Zhang, Alex Hay-Man Ng, Long Tang, Bohua Ling, Dan Wang, Yuelin Hou

Remote Sensing · 23.10.2025 · 17 (21) · 3515

DOI: 10.3390/rs17213515 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 23.10.2025. Verfasst von Qiong Ding und 6 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 66 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Program for Guangdong Introducing Innovative and Entrepreneurial Teams, Guangdong Basic and Applied Basic Research Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Multispectral LiDAR point clouds, which integrate both geometric and spectral information, offer rich semantic content for scene understanding. However, due to data scarcity and distributional discrepancies, existing methods often struggle to balance accuracy and efficiency in complex urban environments. To address these challenges, we propose DA-GSGTNet, a novel segmentation framework that integrates Gated Stratified Graph Transformer Blocks (GSGT-Block) with Dynamic Aggregation Transition Down (DATD). The GSGT-Block employs graph convolutions to enhance the local continuity of windowed attention in sparse neighborhoods and adaptively fuses these features via a gating mechanism. The DATD module dynamically adjusts k-NN strides based on point density, while jointly aggregating coordinates and feature vectors to preserve structural integrity during downsampling. Additionally, we introduce a relative position encoding scheme using quantized lookup tables with a Euclidean distance bias to improve recognition of elongated and underrepresented classes. Experimental results on a benchmark multispectral point cloud dataset demonstrate that DA-GSGTNet achieves 86.43% mIoU, 93.74% mAcc, and 90.78% OA, outperforming current state-of-the-art methods. Moreover, by fine-tuning from source-domain pretrained weights and using only ~30% of the training samples (4 regions) and 30% of the training epochs (30 epochs), we achieve over 90% of the full-training segmentation accuracy (100 epochs). These results validate the effectiveness of transfer learning for rapid convergence and efficient adaptation in data-scarce scenarios, offering practical guidance for future multispectral LiDAR applications with limited annotation.

Förderung

  • National Natural Science Foundation of China
  • Program for Guangdong Introducing Innovative and Entrepreneurial Teams
  • Guangdong Basic and Applied Basic Research Foundation
  • Guangzhou Science and Technology Plan Project

Zitieren

APA

Ding, Q., Zhang, R., Ng, A. H., Tang, L., Ling, B., Wang, D., & Hou, Y. (2025). DA-GSGTNet: Dynamic Aggregation Gated Stratified Graph Transformer for Multispectral LiDAR Point Cloud Segmentation. Remote Sensing, 17(21), 3515. https://doi.org/10.3390/rs17213515

Vancouver

Ding Q, Zhang R, Ng AH, Tang L, Ling B, Wang D, et al. DA-GSGTNet: Dynamic Aggregation Gated Stratified Graph Transformer for Multispectral LiDAR Point Cloud Segmentation. Remote Sensing. 2025;17(21):3515. doi:10.3390/rs17213515

BibTeX
@article{Ding2025,
  author = {Ding, Qiong and Zhang, Runyuan and Ng, Alex Hay-Man and Tang, Long and Ling, Bohua and Wang, Dan and Hou, Yuelin},
  title = {DA-GSGTNet: Dynamic Aggregation Gated Stratified Graph Transformer for Multispectral LiDAR Point Cloud Segmentation},
  journal = {Remote Sensing},
  year = {2025},
  volume = {17},
  number = {21},
  pages = {3515},
  doi = {10.3390/rs17213515}
}
Kostenlos starten →