EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

SPPSFormer: High-Quality Superpoint-Based Transformer for Roof Plane Instance Segmentation from Point Clouds

Cheng Zeng, Xiatian Qi, Huifan Wang, Kai Sun, Pengcheng Zhong, Qiao Xu, Yan Meng, Yangjie Sun, Yuxuan Liu

Remote Sensing · 02.07.2026 · 18 (13) · 2144

DOI: 10.3390/rs18132144 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 02.07.2026. Verfasst von Cheng Zeng und 8 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 96 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Hubei Provincial Key Research and Development Program, Hubei Provincial Major Science and Technology Program, Wuhan Natural Science Foundation Focused Program.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Superpoint Transformers use superpoints as the basic processing units, thereby significantly reducing the number of tokens processed by Transformers. However, they have been seldom employed in point cloud roof plane instance segmentation, and existing superpoint Transformers suffer from limited performance due to the use of low-quality superpoints. To address this challenge, we establish a set of criteria that high-quality superpoints for Transformers should satisfy and introduce a corresponding two-stage superpoint generation process. The superpoints generated by our method not only have accurate boundaries, but also exhibit consistent geometric sizes and shapes, which greatly benefit the feature learning of superpoint Transformers. To compensate for the limitations of deep learning features when the training set size is limited, we incorporate multidimensional handcrafted features into the model. Additionally, we design a decoder that combines a Kolmogorov–Arnold Network with a Transformer module to improve instance prediction and mask extraction. Finally, our network’s predictions are refined using traditional algorithm-based post-processing. For evaluation, we annotated a real-world dataset and corrected annotation errors in the existing RoofN3D dataset. Experimental results show that our method achieves state-of-the-art performance on our dataset, as well as both the original and corrected RoofN3D datasets. Our model also shows significant advantages over existing methods when handling data with low point density, large density variations, or low 3D point precision. Moreover, it is not sensitive to plane boundary annotations during training, significantly reducing the annotation burden. We will release our code, trained models, and datasets.

Förderung

  • Hubei Provincial Key Research and Development Program
  • Hubei Provincial Major Science and Technology Program
  • Wuhan Natural Science Foundation Focused Program
  • Key Laboratory of China-ASEAN Satellite Remote Sensing Applications, Ministry of Natural Resources of the People’s Republic of China
  • Knowledge Innovation Program of Wuhan-Shuguang Project
  • Chinese Academy of Surveying and Mapping
  • National Natural Science Foundation of China
  • State Key Laboratory of Spatial Datum

Zitieren

APA

Zeng, C., Qi, X., Wang, H., Sun, K., Zhong, P., Xu, Q., Meng, Y., Sun, Y., & Liu, Y. (2026). SPPSFormer: High-Quality Superpoint-Based Transformer for Roof Plane Instance Segmentation from Point Clouds. Remote Sensing, 18(13), 2144. https://doi.org/10.3390/rs18132144

Vancouver

Zeng C, Qi X, Wang H, Sun K, Zhong P, Xu Q, et al. SPPSFormer: High-Quality Superpoint-Based Transformer for Roof Plane Instance Segmentation from Point Clouds. Remote Sensing. 2026;18(13):2144. doi:10.3390/rs18132144

BibTeX
@article{Zeng2026,
  author = {Zeng, Cheng and Qi, Xiatian and Wang, Huifan and Sun, Kai and Zhong, Pengcheng and Xu, Qiao and Meng, Yan and Sun, Yangjie and Liu, Yuxuan},
  title = {SPPSFormer: High-Quality Superpoint-Based Transformer for Roof Plane Instance Segmentation from Point Clouds},
  journal = {Remote Sensing},
  year = {2026},
  volume = {18},
  number = {13},
  pages = {2144},
  doi = {10.3390/rs18132144}
}
Kostenlos starten →