Wednesday, May 1, 2013

Procedural Tree Generation

I thought I should post a link to my Honours thesis for posterity, considering it swallowed at least 3 months of my life. My research formed one third of a project to generate a 'realistic' 3D model of a tree from a directed acyclic graph. I focused on generating the mesh for the trunk and branches, while my two teammates researched bark and foliage generation.

The topic required that a single seamless polygon mesh be generated for all of the branches. This allowed a smooth subdivision surface to be derived. Generating a seamless mesh proved a challenging problem, particularly at the points where branches met. I ended up solving this by constructing a 3D convex hull around the point where the branches met. In hindsight I probably should have used adaptive distance fields instead.

No comments:

Post a Comment