A Fast Shape Reconstruction Method for Large-Scale Terrain Point Cloud

doi:10.3795/KSME-A.2025.49.11.879

A Fast Shape Reconstruction Method for Large-Scale Terrain Point Cloud

2025-11-01

0

PRE

AI

DOI：10.3795/KSME-A.2025.49.11.879

原文链接

原文求助

摘要

En 中文

The reconstruction of three-dimensional shapes from point cloud data has been extensively applied across diverse domains. Conventional methods, such as Delaunay triangulation, voxel-based reconstruction, and Poisson surface reconstruction, often incur significant computational costs when processing large-scale point clouds and are prone to reduced accuracy in the presence of noise and irregular point distributions. This study introduces a mesh offset method tailored for large-scale terrain point cloud reconstruction. The approach employs a multi-level grid structure to efficiently suppress clustered noise. A two-dimensional grid encompassing the point cloud is generated, and the mean elevation within each cell is computed. Points that deviate significantly from the mean are iteratively removed, facilitating robust reconstruction even for highly scattered point distributions. Displacement values are then calculated for each grid node, which are subsequently adjusted along the z-axis to generate the final mesh. The results demonstrate that the proposed method achieves efficient and reliable reconstruction of large-scale terrain point clouds acquired via LiDAR, with markedly reduced computation time.

A Fast Shape Reconstruction Method for Large-Scale Terrain Point Cloud

摘要

期刊

机构

引用论文