Peppermint

Goal of the project

Given a point cloud representing fragments:

• identify contiguous regions of points which belong to the same fragment based on neighborhood criteria
• extract surfaces from each fragment
• mesh these surfaces with a suitable triangulation method
• have a basic mock simulation of fragment travel and break up
• implement an algorithm which efficiently updates the:
  • fragment list (particle / fragment membership)
  • surface mesh (node positions)
  • surface topology

I use this project to teach myself C and a bunch of data structures and algorithms, but I'm (still) optimistic that the end result will implement the goals above well enough for production code.

I called it peppermint because of the large quantity of peppermint gum consumed while working on this project.

Features

• Not many, so far

TO DO

• finish implementing basic helper function for sorting, vector math, memory management
• decide on a data structure for the particles
  • probably struct of long vectors instead of vector of structs for sorting
  • consider potential addition of bonds in this decision
• add shape generation which fills various geometric shapes with particles for testing purposes later on
• implement various data structures and algorithms for efficient neighborhood search etc
  • k d tree
  • i j k search grid
  • marching cubes
  • octree
  • flood fill
  • other voxel based methods
• output writer and python script for results visualization
• identification of contiguous fragments in particle cloud
• vtk output writer for results visualization
• implement marching cubes for surface construction on fragments
• use octree chunk elimination and flood fill to discard inner surfaces
• implement poisson disk sampling and GreedyProjectionTriangulation for surface (re)meshing
• worry about my sanity, probably
• add mock particle simulation
  • something like two bananas crashing into each other and "fragmenting" should provide sufficient complexity for testing
• write algorithm for adaptive
  • fragment updates
  • surface mesh modification
  • surface mesh reconstruction
• add OMP and SIMD parallelization
• potentially look into CUDA as well
• explore PCL and see how it compares

License

All rights reserved.