Real-time Surface Rendering for Huge Point Clouds

PR, DA

Stefan Ohrhallinger, Michael Wimmer

Content:

Description

Scans of archaeological sites generate huge point clouds. Such data do not contain a surface, so rendering just the points has many problems, such as visibility ('see-through' walls), unavailability of physics-based properties for e.g. collision detection, or slow rendering because of too many points. Doing ray-casting instead enables real-time surface construction as approach to solve these problems.

Task

Implement k-nearest neighbor search in a given multi-resolution data structure (octree with nested grids) for ray-cast points in order to render their umbrella of triangles as approximate surface.
Compare efficiency of knn search and quality of umbrella creation with existing work.

Requirements

Reasonable C++ knowledge. Strong interest in algorithms. Ability to visualize 3D data structures well.

Environment

The XGRT graphics framework from MPII will be used to pre-process the point sets. It is written in C++ and can be compiled in Visual Studio 2008. Developing under Linux is encouraged.