ThreeCrate Reconstruction

Surface reconstruction algorithms for generating meshes from point clouds.

Features

• Poisson Reconstruction: High-quality surface reconstruction using Poisson solving
• Ball Pivoting: Fast surface reconstruction for uniformly sampled point clouds
• Alpha Shapes: Geometric reconstruction using alpha complex
• Delaunay Triangulation: 2D/3D triangulation for mesh generation
• Parallel Processing: Multi-threaded algorithms for large datasets

Algorithms

Poisson Surface Reconstruction

• High-quality mesh generation from oriented point clouds
• Handles noise and irregular sampling well
• Produces watertight meshes
• Configurable octree depth and sample density

Ball Pivoting Algorithm (BPA)

• Fast reconstruction for uniformly sampled point clouds
• Good for dense, noise-free data
• Preserves sharp features and boundaries
• Configurable ball radius and clustering

Alpha Shapes

• Geometric approach using alpha complex
• Good for shape analysis and boundary detection
• Multiple levels of detail with different alpha values
• Handles complex topologies

Usage

Add this to your Cargo.toml:

[dependencies]
threecrate-reconstruction = "0.1.0"
threecrate-core = "0.1.0"

Example

use threecrate_reconstruction::{poisson_reconstruction, ball_pivoting_reconstruction};
use threecrate_core::{PointCloud, Point3f};

// Load point cloud with normals
let cloud = PointCloud::from_points(vec![/* points */]);

// Poisson reconstruction
let mesh = poisson_reconstruction(&cloud, 6, 1.0, 0.1)?;
println!("Generated mesh with {} faces", mesh.faces.len());

// Ball pivoting reconstruction
let radius = 0.1;
let mesh = ball_pivoting_reconstruction(&cloud, radius)?;
println!("Generated mesh with {} faces", mesh.faces.len());

Algorithm Details

Poisson Parameters

• Octree Depth: Controls mesh resolution (6-10 recommended)
• Sample Density: Point density factor (0.5-2.0)
• Confidence Threshold: Quality filtering (0.0-1.0)

Ball Pivoting Parameters

• Ball Radius: Reconstruction radius (depends on point density)
• Clustering: Remove duplicate vertices and faces
• Normal Consistency: Ensure consistent face orientations

Requirements

• Point clouds with estimated normals for best results
• Sufficient point density for reconstruction
• Reasonable memory for large point clouds

License

This project is licensed under either of

• Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.