EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

3D Geometry-Aware Efficient Feature Matching for Weakly Textured Scenes

Libo Sun iD, Yidong Yan iD, Wenqi Yang, Wenhu Qin

Journal of Imaging · 07.06.2026 · 12 (6) · 253

DOI: 10.3390/jimaging12060253 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Imaging, online erschienen am 07.06.2026. Verfasst von Libo Sun und 3 weiteren Autor:innen. Beteiligte Einrichtung: School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China. Die Arbeit nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: the Key R&D Program of Jiangsu Province, Basic Research Program of Jiangsu, Advanced Technology Research and Development Program of Jiangsu.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Local feature matching plays a critical role in robotic SLAM and visual localization. However, in weakly textured indoor industrial environments, lightweight appearance-based methods often struggle to learn discriminative and stable local features. To address this challenge, this paper proposes GAEFeat, short for Geometry-Aware Efficient Feature, a lightweight vision–geometric feature learning network. To address the scarcity of specialized training data, we integrated robotic arm pose priors with depth information to automatically generate cross-view supervision signals and surface-normal labels. Based on this strategy, we constructed two complementary datasets, including a simulated dataset and a real-world dataset, to support feature learning and evaluation in weakly textured indoor industrial environments. For feature extraction, we design a dual enhancement mechanism consisting of a geometric auxiliary branch and a geometry-aware enhancement (GAE) module. The former guides the network to perceive local surface structures through surface normal supervision, while the latter utilizes a gating mechanism to achieve deep fusion between geometric priors and 2D texture descriptors. Experimental results demonstrate that GAEFeat achieves strong robustness and high inference efficiency in relative pose estimation, homography estimation, and visual localization tasks, with particularly notable advantages in near-field, weakly textured industrial scenes. The framework achieves an inference latency of only 3.9 ms on the NVIDIA Jetson AGX Orin edge platform, demonstrating its real-time capability and practical potential for deployment in edge computing environments.

Schlagwörter

Stichwörter der Autor:innen

RGB-D geometryedge computingfeature matchingindustrial visionkinematics supervisionweak texture

Autor:innen und Institutionen

  • School of Instrument Science and Engineering, Southeast University, Nanjing 210096, China.

ORCID-Profile: Libo Sun, Yidong Yan

Förderung

  • the Key R&D Program of Jiangsu Province
  • Basic Research Program of Jiangsu
  • Advanced Technology Research and Development Program of Jiangsu
  • the Key R&D Program of Jiangsu Province

Zitieren

APA

Sun, L., Yan, Y., Yang, W., & Qin, W. (2026). 3D Geometry-Aware Efficient Feature Matching for Weakly Textured Scenes. Journal of Imaging, 12(6), 253. https://doi.org/10.3390/jimaging12060253

Vancouver

Sun L, Yan Y, Yang W, Qin W. 3D Geometry-Aware Efficient Feature Matching for Weakly Textured Scenes. Journal of Imaging. 2026;12(6):253. doi:10.3390/jimaging12060253

BibTeX
@article{Sun2026,
  author = {Sun, Libo and Yan, Yidong and Yang, Wenqi and Qin, Wenhu},
  title = {3D Geometry-Aware Efficient Feature Matching for Weakly Textured Scenes},
  journal = {Journal of Imaging},
  year = {2026},
  volume = {12},
  number = {6},
  pages = {253},
  doi = {10.3390/jimaging12060253}
}
Kostenlos starten →