Inhalt

•  
   Acknowledgments
•  
   Abstract
•  
   Contents
•  
   1 Introduction
•  
   2 Perception of the Vista Space
•  
   2.1 Definition of the Vista Space
•  
  2.2 BIRON – the BIelefeld Robot companiON
•  
   2.2.1 The Robot Platform
•  
   2.2.2 The ``Home Tour'' Scenario
•  
   2.2.3 Vista Space Scenes in the ``Home Tour''
•  
  2.3 A Sensor for Perceiving Spatial Structures in 3D
•  
   2.3.1 Working Principle of the SwissRanger Camera
•  
   2.3.2 Preprocessing of SwissRanger Data
•  
  2.4 Basic Processing of a Single Percept
•  
   2.4.1 Computing Oriented Particles
•  
   2.4.2 Extracting Planar Surfaces
•  
  2.5 Basic Processing of Consecutive Percepts
•  
   2.5.1 Extending 3D Data with Velocities
•  
   2.5.2 Fusing Sets of Point Clouds
•  
   3 Learning Holistic Scene Models from Spatial Layouts
•  
   3.1 Motivation
•  
  3.2 Related Work
•  
   3.2.1 From Robotics Perspective
•  
   3.2.2 From Vision Perspective
•  
   3.2.3 Approaches Chosen for Comparison
•  
   3.2.4 Contribution of the Holistic Scene Model
•  
  3.3 The Holistic Scene Representation
•  
   3.3.1 The Scene Descriptor from 3D Data
•  
   3.3.2 The Scene Descriptor from 2D Data
•  
   3.3.3 Training Room Models and Combining Single Classifications
•  
  3.4 Evaluation
•  
   3.4.1 The 3D Indoor Database
•  
   3.4.2 Classifier Selection and Training
•  
   3.4.3 Classification Performance for Different Window Sizes
•  
   3.4.4 Classification Performance per Class
•  
   3.4.5 Feature Concatenation vs. Classifier Fusion
•  
   3.4.6 Room Label Distribution along Example Sequences
•  
   3.4.7 Correlations between Sub-Vectors of the 3D Feature Vector
•  
   3.5 Conclusion and Outlook
•  
   4 Learning Aligned Scene Models from Spatial Descriptions
•  
   4.1 Motivation
•  
  4.2 Related Work
•  
   4.2.1 Scene Interpretation from Verbal Input
•  
   4.2.2 Scene Interpretation from Visual Input
•  
   4.2.3 Integration of Verbal and Visual Scene Interpretations
•  
   4.2.4 Contribution of the Aligned Scene Model
•  
   4.3 Empirical Analysis of Spatial Scene Descriptions
•  
   4.4 The Computational Model
•  
   4.4.1 From Verbal Descriptions to Set of Trees
•  
   4.4.2 Inferring Initial 3D Scene Structures
•  
   4.4.3 Adapting the Initial Scene Structures to the Visual Perception
•  
  4.5 Evaluation
•  
   4.5.1 Analysis of an Example Model
•  
   4.5.2 Analysis of Level-1 Structures
•  
   4.5.3 Analysis of Level-2 Structures
•  
   4.5.4 Influence of Object Detection Errors on Model Formation
•  
   4.6 Conclusion and Outlook
•  
   5 Learning Articulated Scene Models from Spatial Changes
•  
   5.1 Motivation
•  
  5.2 Related Work
•  
   5.2.1 Detection of Moving Objects and Static Scene Modeling
•  
   5.2.2 Detection of Movable Objects and Semantic Areas
•  
   5.2.3 Contribution of the Articulated Scene Model
•  
   5.3 The Analysis of a Dynamic Scene
•  
   5.3.1 Entity Tracking
•  
   5.3.2 Static Background Adaptation and Movable Object Detection
•  
  5.4 Evaluation
•  
   5.4.1 Qualitative Evaluation of a Test Sequence
•  
   5.4.2 Quantitative Evaluation of a Set of Test Sequences
•  
  5.5 Applications of the Articulated Scene Model
•  
   5.5.1 Object Segmentation
•  
   5.5.2 Model Propagation from View to View
•  
   5.5.3 Object Articulation
•  
   5.6 Conclusion and Outlook
•  
   6 Summary
•  
  A Appendix – Scene Classification
•  
   A.1 3D Indoor Scene Categorization – A Prove of Concept
•  
   A.2 Equivalence of Form Factors for 2D Boxes
•  
   B Appendix – Scene Descriptions
•  
   B.1 Pilot Study: Playroom
•  
   B.2 Main Study: Playroom
•  
   B.3 Main Study: Living Room
•  
   Bibliography
•  
   List of Figures
•  
   List of Tables
•  
   Acronyms