nnu-gisa/surface-reconstruction

Repo of code written for surface reconstruction from point clouds

Surface Reconstruction

Repo of code written for surface reconstruction from point clouds

Pipeline

1. Input: unoriented point clouds (xyz values, and optionally RGB values in .ply format)
2. Optionally, chunk ply and use the smaller chunks of ply to run step 3 and 4 on different nodes on a computing cluster. Use ChunkPLY provided by KBH06.
   ./ChunkPLY --in input-filename.ply --out output-filename.chunks --width 200
3. Normal estimation: Needed because unless the point cloud has very high sampling density and is noise-free, which Delaunay-based reconstruction methods would work fine, reliable normals and oriented normals especially at each point are crucial for surface reconstruction. Use ComputeNormals.py to add normals information to .ply file if it does not contain normals.
4. Use Poisson Reconstruction algorithm KBH06 at depth => 10 (14 preferred).
   ./PoissonRecon --in input-filename.ply --out output-filename.ply --depth 14
5. Evaluate the resulting mesh model in MeshLab or MeshMixer