EUVIMEDEUVIMED entdecken
Crossref

An adaptive curvature–residual weighting method for tooth point cloud registration

Jingchen Xie, Guoliang Chen

Measurement Science and Technology · 10.04.2026 · 37 (15) · 156306

DOI: 10.1088/1361-6501/ae5cc4 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 16.04.2026. Verfasst von Jingchen Xie und 1 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To meet the stringent requirements for accuracy and robustness in fine tooth point cloud registration for digital dentistry and surgical navigation, this paper proposes an adaptive curvature–residual weighted registration method for tooth point clouds. The core contribution of this work is the introduction of curvature-based and residual-based weighting as two complementary components that jointly balance convergence speed and stability. Conventional iterative closest point (ICP) and its variants often suffer from unstable convergence and entrapment in local minima when confronted with pronounced geometric feature variations, noise contamination, and mismatched point correspondences. To address these limitations, a weighting mechanism is constructed from two complementary perspectives: geometric saliency and matching reliability. At the geometric feature level, the local neighborhood covariance structure of the tooth point cloud is analyzed to derive a curvature-based feature weighting function, which strengthens the registration constraints imposed by high-curvature regions such as cusps, ridges, and grooves. At the matching error level, a residual weighting strategy based on robust kernel functions is introduced to adaptively suppress point pairs with large residuals, thereby reducing the adverse effects of noise and mismatches on pose estimation. By jointly incorporating curvature weights and residual weights into a weighted ICP optimization framework, a curvature–residual collaboratively driven registration model is established. Experimental results on multiple real tooth point cloud datasets demonstrate that, in most test scenarios, the proposed method achieves clear improvements in both registration accuracy and convergence stability compared with conventional ICP. Moreover, superior registration performance is observed on tooth models with prominent geometric features, validating the effectiveness and practical applicability of the proposed approach for complex oral point cloud registration tasks.

Zitieren

APA

Xie, J., & Chen, G. (2026). An adaptive curvature–residual weighting method for tooth point cloud registration. Measurement Science and Technology, 37(15), 156306. https://doi.org/10.1088/1361-6501/ae5cc4

Vancouver

Xie J, Chen G. An adaptive curvature–residual weighting method for tooth point cloud registration. Measurement Science and Technology. 2026;37(15):156306. doi:10.1088/1361-6501/ae5cc4

BibTeX
@article{Xie2026,
  author = {Xie, Jingchen and Chen, Guoliang},
  title = {An adaptive curvature–residual weighting method for tooth point cloud registration},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {15},
  pages = {156306},
  doi = {10.1088/1361-6501/ae5cc4}
}
Kostenlos starten →