Towards semantic scene analysis with time-of-flight cameras

Authors:
Dirk Holz;Ruwen Schnabel;David Droeschel;Jörg Stückler;Sven Behnke
Affiliations:
University of Bonn, Institute of Computer Science VI;University of Bonn, Institute of Computer Science II;University of Bonn, Institute of Computer Science VI;University of Bonn, Institute of Computer Science VI;University of Bonn, Institute of Computer Science VI
Venue:
RoboCup 2010
Year:
2011

Citing 5
Cited 3

MLESAC: a new robust estimator with application to estimating image geometry

Computer Vision and Image Understanding - Special issue on robusst statistical techniques in image understanding
Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography

Communications of the ACM
Towards semantic maps for mobile robots

Robotics and Autonomous Systems
Three-dimensional mapping with time-of-flight cameras

Journal of Field Robotics - Three-Dimensional Mapping, Part 2
Close-range scene segmentation and reconstruction of 3D point cloud maps for mobile manipulation in domestic environments

IROS'09 Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems

Efficient 3D object perception and grasp planning for mobile manipulation in domestic environments

Robotics and Autonomous Systems
Real-time plane segmentation using RGB-D cameras

Robot Soccer World Cup XV
Plane-based object categorisation using relational learning

Machine Learning

Quantified Score

Hi-index	0.00

Visualization

Abstract

For planning grasps and other object manipulation actions in complex environments, 3D semantic information becomes crucial. This paper focuses on the application of recent 3D Time-of-Flight (ToF) cameras in the context of semantic scene analysis. For being able to acquire semantic information from ToF camera data, we a) pre-process the data including outlier removal, filtering and phase unwrapping for correcting erroneous distance measurements, and b) apply a randomized algorithm for detecting shapes such as planes, spheres, and cylinders. We present experimental results that show that the robustness against noise and outliers of the underlying RANSAC paradigm allows for segmenting and classifying objects in 3D ToF camera data captured in natural mobile manipulation setups.