This repository contains an implementation of the Quickhull algorithm for generating 3d convex hulls from point clouds, written for Unity. A convex hull is the smallest convex shape that contains a given set of points (basically: if you have a bunch of points in space, the convex hull is what you’d get if you “shrinkwrapped” the points as tightly as possible).
The code is commented and documented in the source file, and it should be pretty readable. It is suitable for use in production.
Let me know if you find a use for this, I would love to hear about it!
Convex hulls are useful in a variety of algorithms in graphics and gamedev. The simplest of which is maybe as a low-poly “rock generator”, which is included in the project.
The way this works is pretty simple: just generate a bunch of random points in the unit sphere, use the ConvexHullCalculator to calculate convex hulls for them which is used to create meshes.
The only file that really matters is ConvexHullCalculator.cs in the Scripts folder. You can either just copy that file into your project, or add the repository into a Plugins folder.
ConvexHullCalculator.cs defines a class called ConvexHullCalculator that can calculate 3d convex hulls from a point cloud containing at least 4 points. It returns lists of vertices, triangles and normals that can be directly converted into a mesh (see example for how to do this).
If there are less than 4 points, or if all points are coplanar, the calculator will throw an exception, as 3d convex hulls aren’t really defined in those cases.
The class uses a number of internal buffers for doing the calculation, but they are all reused for subsequent calculations in order to avoid generating garbage. You can see in the example that the same ConvexHullCalculator (as well as the List<T>’s used to store the results) can be reused many times so you can avoid generating garbage.
The calculator itself doesn’t interact at all with the Unity engine, so it’s perfectly safe to run the calculations on background threads. However, the calculator is not thread-safe, so a single calculator cannot be used at the same time by two or more threads. If you wish to run multiple calculations in parallel, use one calculator per thread. Note also that the calculator is entirely serial, which means that a single calculator will max out at most one logical core.