Outline
Motivation
Sensor Systems
Depth Reconstruction from Stereo
Disparity
Error Sources in the Stereo 3D-Reconstruction
Reduction of the complexity
Correspondence Problem
Epipolar-Geometry
Feature Detection
Influence of the Orientation Error
Transformation to the Parallel-Cameras Case
Dynamic Local Map (DLM)
DLM - Principle of Geometrical Fusion
DLM - Structure for fast, efficient storage of a local area
DLM - Filtering
DLM - Stabilization
Re-calibration based on the DLM
Re-calibration of the external parameters
Results of the on-line Re-calibration
Abstraction of knowledge within the system
Estimation of the free space
Navigation map
Path planing with diffusion analogy
Optical Bumper
Support for topological modeling
Initial Localization
Results (Exploration)
Results (3D-Reconstruction)
Conclusion
Future Work
Email: burschka@cs.jhu.edu
Home Page: http://www.cs.jhu.edu/~burschka