Abstract:We present a novel approach for generating isotropic surface triangle meshes directly from unoriented 3D point clouds, with the mesh density adapting to the estimated local feature size (LFS). Popular reconstruction pipelines first reconstruct a dense mesh from the input point cloud and then apply remeshing to obtain an isotropic mesh. The sequential pipeline makes it hard to find a lower-density mesh while preserving more details. Instead, our approach reconstructs both an implicit function and an LFS-aware mesh sizing function directly from the input point cloud, which is then used to produce the final LFS-aware mesh without remeshing. We combine local curvature radius and shape diameter to estimate the LFS directly from the input point clouds. Additionally, we propose a new mesh solver to solve an implicit function whose zero level set delineates the surface without requiring normal orientation. The added value of our approach is generating isotropic meshes directly from 3D point clouds with an LFS-aware density, thus achieving a trade-off between geometric detail and mesh complexity. Our experiments also demonstrate the robustness of our method to noise, outliers, and missing data and can preserve sharp features for CAD point clouds.

What problem does this paper attempt to address?

This paper attempts to solve the problem of generating isotropic surface triangular meshes with Local Feature Size (LFS) awareness from unoriented 3D point clouds. Specifically, existing reconstruction pipelines usually first reconstruct a dense mesh from the input point cloud and then obtain an isotropic mesh through remeshing. This sequential processing method makes it difficult to find a lower - density mesh while maintaining more details. In addition, obtaining these adaptive isotropic meshes requires a delicate balance between mesh complexity and reconstruction accuracy. To solve these problems, the authors propose a new method that directly reconstructs the implicit function and the LFS - aware mesh size function simultaneously from the unoriented 3D point cloud and uses this information to generate the final LFS - aware mesh without remeshing. This method directly estimates the LFS from the input point cloud by combining the local radius of curvature and the shape diameter and proposes a new mesh solver to solve an implicit function whose zero - level set defines the surface and does not require normal directions. ### Main contributions: 1. **LFS Estimation**: A new method is proposed to directly estimate the Local Feature Size (LFS) from 3D point clouds, avoiding the method of constructing the medial axis through the Voronoi diagram, which is sensitive to noise and sampling conditions. The new method estimates the LFS by jet fitting and analyzing the Lipschitz distance function along random rays, which is reliable and robust to low noise and non - uniform sampling. 2. **Implicit Function Solving**: A new method is introduced to solve the signed implicit function, which is mainly divided into three steps: - Generate and discretize a multi - domain, which consists of a thin envelope around the input points and the complement of the loose bounding sphere of the input points. - Solve the signed implicit function from the sign guesses of the edges by the least - squares method, which can resist outliers and fill large holes. - Construct a robust signed implicit function to provide robustness to noise. 3. **Experimental Verification**: Extensive experiments have been carried out on synthetic and real - world 3D point clouds, including the AIM@Shape dataset and the ABC dataset. The experimental results show that the output mesh generated by this method is valid (i.e., no intersection), has an adaptive mesh size, and has a smaller reconstruction error than previous methods. ### Method Overview: 1. **LFS Estimation**: - **Local Radius of Curvature**: The local radius of curvature is estimated by local jet fitting, and the jet fitting constructs a local Monge coordinate system whose basis is defined by the maximum principal curvature, the minimum principal curvature, and the normal direction. - **Shape Diameter**: An extended version of the shape diameter function is estimated by the double - cone search algorithm, which takes into account local thickness and separation. 2. **Implicit Function Solving**: - **Multi - Domain Discretization**: Construct a multi - domain containing a thin envelope of the input points and a loose bounding sphere. - **Sign Guessing and Data Fitting**: Solve the signed implicit function from the sign guesses of the edges by the least - squares method. - **Robust Distance Function**: Use the binary sign values of the signed implicit function to symbolize the unsigned robust distance function, thereby obtaining a new implicit function. 3. **LFS - Aware Meshing**: - Use the Delaunay refinement meshing method to extract the final adaptive isotropic mesh, which controls the refinement termination according to the LFS - aware mesh size function. ### Formula Representation: - **LFS Estimation**: \[ \text{lfs}(x)=\min\left( r(x), 0.5\cdot\delta(x) \right) \] where \( r(x) \) represents the local radius of curvature and \(\delta(x)\) represents the shape diameter function. - **Local Radius of Curvature**: \[ r(x)=\frac{1}{|h(x)|} \] where \( |h(x)| \) is the maximum absolute value of the curvature estimated by jet fitting. - **Shape Diameter**: \[ \delta(x)=\min\left( \tau(x), \sigma(x) \right) \] where \( \tau(

LFS-Aware Surface Reconstruction from Unoriented 3D Point Clouds

A clustering approach to free form surface reconstruction from multi-view range images

Adaptive Mesh Reconstruction of Point Cloud with Feature Preserved

Improved Fast 3D Reconstruction Algorithm Based on Laser Scanning Point Cloud

A Novel Pointcloud-Based Isosurface Extraction

A Two-Phase Approach of Progressive Mesh Reconstruction from Unorganized Point Clouds

A Mesh Reconstruction Algorithm Driven by an Intrinsic Property of a Point Cloud.

Mesh Reconstruction of Noisy Point Cloud Based on Intrinsic Property

A Fast Progressive Surface Reconstruction Algorithm for Unorganized Points

3D Reconstruction Approach for Outdoor Scene Based on Multiple Point Cloud Fusion

An Efficient Method to Mesh Point Cloud

Multi-scale Surface Reconstruction Based on a Curvature-Adaptive Signed Distance Field.

Learning Unsigned Distance Fields from Local Shape Functions for 3D Surface Reconstruction

Plane Fitting in 3D Reconstruction to Preserve Smooth Homogeneous Surfaces

A Novel Surface Reconstruction Framework from Point Clouds Without Normal

Point normal orientation and surface reconstruction by incorporating isovalue constraints to Poisson equation

Direct Sampling on Surfaces for High Quality Remeshing

Reconstruction of tubular structures from 2.5D point clouds: A mesophotic gorgonian coral case study

Fan-Meshes: a Geometric Primitive for Point-Based Description of 3D Models and Scenes

Multisource Point Clouds, Point Simplification and Surface Reconstruction

Reconstructing Surfaces for Sparse Point Clouds with On-Surface Priors