Digital Elevation Mapping using stereoscopic vision

navigational decisions can be based. I will present the results of a study into stereo
vision aimed at the development of a method for three-dimensional mapping of outdoor
scenes. The acquired map covers a local environment and is extended incrementally as
the robot moves. Areas covered in this talk are:

- Disparity computation (as means for local map building)
- Partial map registration / motion from stereo
- The octree as a data structure for 3D occupancy grid representation
- 3D- model extraction from 3D occupancy grid

Together these parts make up an algorithm for incremental 3D-map building from
stereo-sequences. Advantages and drawbacks are discussed as well.