ThreeCrate Reconstruction

Surface reconstruction algorithms for generating meshes from point clouds.

Algorithms

• Poisson Surface Reconstruction: Watertight meshes from oriented point clouds; configurable octree depth
• Ball Pivoting Algorithm (BPA): Fast reconstruction for uniformly sampled clouds; multi-scale and adaptive radius
• Alpha Shapes: Non-convex surface extraction via alpha complex; multiple levels of detail
• Delaunay Triangulation: Complete 3D triangulation with multiple projection methods
• Marching Cubes: Volumetric/implicit surface to mesh conversion
• Moving Least Squares (MLS): Smooth surface fitting with multiple weight functions
• Unified Pipeline: auto_reconstruct() — automatic algorithm selection based on point cloud characteristics, with quality metrics and validation

Usage

Add this to your Cargo.toml:

[dependencies]
threecrate-reconstruction = "0.6.0"
threecrate-core = "0.6.0"

Example

use threecrate_reconstruction::{
    auto_reconstruct,
    poisson_reconstruction, poisson_reconstruction_default,
    ball_pivoting_reconstruction, BallPivotingConfig,
};
use threecrate_core::PointCloud;

// Automatic algorithm selection
let mesh = auto_reconstruct(&cloud)?;
println!("Generated mesh: {} faces", mesh.faces.len());

// Poisson reconstruction
let mesh = poisson_reconstruction_default(&normals_cloud)?;

// Ball pivoting with custom config
let config = BallPivotingConfig { radius: 0.1, ..Default::default() };
let mesh = ball_pivoting_reconstruction(&cloud, config)?;

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)
• Multi-scale: Adaptive radius selection for varying densities

Requirements

• Point clouds with estimated normals produce the best results with Poisson and MLS
• Sufficient point density relative to the ball radius for BPA

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.