Point Cloud Registration for Lava Tube Surface Reconstruction Using Curvature-Optimized Projection

Jiaqi Yao, Yanqiu Wang, Wen Li, Han Yuan, Hao Ai, Zhenguo Ji, Fu Zheng, Zhibin SunIEEE Access 2025 · DOI

Abstract

The surface reconstruction of point clouds is a key technology in lunar exploration. The generation of high-precision three-dimensional models of lunar lava tubes via surface reconstruction provides technical support for subsequent exploration. To address the problem of detailed feature loss in the surface reconstruction of natural lava tubes, our study proposed a curvature-optimized point cloud surface reconstruction method. First, the normal vectors were estimated and normalized on an unstructured point cloud using Principal Component Analysis. Subsequently, the point cloud was projected using the Moving Least Squares method based on an anisotropic weighting function to obtain curvature-optimized point cloud data. Finally, implicit surface reconstruction and triangular mesh construction were achieved using the Poisson algorithm. Our study analyzed the morphological parameters of the Indian lava tube and investigated the impact of the parameter settings on registration, resampling, and surface reconstruction. The results show that our method can achieve resampling and surface reconstruction with Root Mean Square Error of less than 1.4 cm and 6.8 cm, respectively. Compared with Farthest Point Sampling, Normal Space Sampling, Uniform Sampling, and Voxel Sampling methods, our method has a higher reconstruction surface accuracy and is suitable for high-precision surface reconstruction of natural lava tube point clouds. It can be used to construct high-precision three-dimensional models of the lunar lava tubes.
BibTeX
@inproceedings{yao2025point,
  author = {Jiaqi Yao and Yanqiu Wang and Wen Li and Han Yuan and Hao Ai and Zhenguo Ji and Fu Zheng and Zhibin Sun},
  title = {{Point Cloud Registration for Lava Tube Surface Reconstruction Using Curvature-Optimized Projection}},
  booktitle = {IEEE Access},
  year = {2025},
  doi = {10.1109/access.2025.3550497},
}

Loading related content...

Figure