3D Reconstruction for Exploration of Indoor Environments
Autonomous exploration of arbitrary indoor environments with a mobile robot depends on a reliable self-localization strategy. Existing approaches that use only 2D distance information from e.g. planar laser scanners may fail in highly cluttered areas due to the lack of stable landmark detection. This paper presents an approach for extracting room and furniture primitives from a 3D point cloud by matching shape primitives to the data samples. These basic building blocks can serve as landmarks for relocalization and give hints for interesting places during environmental exploration. Input data is acquired by a tiltable 2D laser scanner in reality and a realistic virtual sensor simulation. In the paper the complete process from sensor data acquisition, data filtering, RANSAC1 based plane extraction and smoothing is described and tested in simulation and reality.
Unable to display preview. Download preview PDF.
- 1.M. A. Fischler and R. C. Bolles. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Commun. ACM 24, 1981.Google Scholar
- 2.J. Weingarten. Feature-Based 3D SLAM. PhD thesis, EPFL, September 2006.Google Scholar
- 3.J. Weingarten, G. Gruener, and R. Siegwart. A fast and robust 3d feature extraction algorithm for structured environment reconstruction. In Proc. of ICAR 2003.Google Scholar
- 4.D. Schmidt, T. Luksch, J. Wettach, and K. Berns. Autonomous behavior-based exploration of office environments. In Proc. of ICINGO 2006.Google Scholar
- 5.R. Schnabel, R. Wahl, and R. Klein. Shape detection in point clouds. Technical report, Computer Graphics Technical Report, 2006.Google Scholar
- 6.T. Braun, J. Wettach, and K. Berns. A customizable, multi-host simulation and visualization framework for robot applications. In Proc. of ICAR 2007.Google Scholar
- 7.X. Astigarraga. 3d reconstruction of structured indoor environments. Master’s thesis, University of Kaiserslautern, March 2007.Google Scholar