Inhalt

•  
   Abstract
•  
   Kurzfassung
•  
   Acknowledgments
•  
   Contents
•  
   List of Abbreviations
•  
  Introduction
•  
   1.1Motivation
•  
   1.2Problem Description
•  
   1.3Key Contribution
•  
   1.4Thesis Outline
•  
  Analysis of 2D/3D Image Data
•  
   2.1 3D Range Measurement
•  
   2.1.1Stereoscopic Imaging
•  
   2.1.2Structured Light Approach
•  
   2.1.3Laser Pulse Range Finder
•  
  2.1.4Time of Flight Camera
•  
   ZESS-Time of Flight Camera10
•  
   2.2 2D/3D Vision System
•  
   2.3The MultiCam
•  
   2.3.1Time of Flight Range Analysis
•  
   2.3.2Range Calibration
•  
   2.3.32D/3D Synchronization
•  
   2.3.42D/3D Image Calibration and Registration
•  
   2.4Summary
•  
   2D/3D Object Recognition
•  
  3.1Feature Extraction
•  
   3.1.1Principal Component Analysis
•  
   Computational Cost of PCA
•  
   Eigenimages and PCA Reconstruction
•  
   3.1.2Linear Discriminant Analysis
•  
   3.1.3Knowledge Based Features
•  
   An Example of Knowledge Based Features using 2D/3D Images
•  
   3.1.4Haar-like Features
•  
   3.2Multimodal Image Segmentation
•  
   3.2.1Range Segmentation
•  
   3.2.2Multimodal Data Fusion
•  
  3.2.3Unsupervised Clustering
•  
   K-Means
•  
   Expectation Maximization
•  
   K-means Expectation Maximization (KEM)
•  
   3.2.4Experiments and Results
•  
   Low Resolution Hand Segmentation Results
•  
   2D/3D Object Segmentation Results
•  
   Integration of Edge Detection and Clustering in 2D/3D Image Segmentation
•  
   3.3Object Classification
•  
   3.3.1Support Vector Machines
•  
   Linear Support Vector Machines - Separable Case (Hard-Margin Classifier)
•  
   Linear Support Vector Machines - Inseparable Case (Soft-Margin Classifier)
•  
   Nonlinear Support Vector Machines - Kernel Based
•  
  3.3.2Moving Object Classification Using Support Vector Machines
•  
   Set-Up and Image Acquisition
•  
   Classification Algorithm
•  
   Classification Results Using PMD TOF Images
•  
   Classification Results Using Stereo Range Images
•  
   3.3.3AdaBoost Classification
•  
  3.3.4 2D/3D Object Detection Using the Viola-Jones Method
•  
   Viola-Jones Method
•  
   Overview of Object Detection Algorithm
•  
   3.4Summary
•  
   2D/3D Object Tracking
•  
   4.1Dynamic Scene Analysis
•  
  4.1.1Background Subtraction
•  
   Adaptive Gaussian Mixture Model
•  
   Background Subtraction Using Range Thresholding
•  
   Evaluation of Results
•  
   4.1.2Real Time Aspects
•  
  4.2Object Representation and Identification
•  
   4.2.1Feature Extraction and Correspondence Matching
•  
  4.2.2Occlusion Handling
•  
   Object Occlusion Detection
•  
   Object Split Detection
•  
   Object Matching
•  
  4.2.3Tracking with Classifiers
•  
   Hand Detection and Tracking in 2D/3D Videos
•  
   Evaluation of Hand Detection Results
•  
   4.3Probabilistic Object Tracking
•  
   4.3.1The Kalman Filter
•  
   4.3.2The CONDENSATION Algorithm
•  
   4.3.3Evaluation of Results
•  
   4.4Summary
•  
  Applications
•  
   5.1Personnel Safety in a Human Robot Cooperation
•  
   5.1.1Background
•  
   5.1.2Dynamic Visual Monitoring
•  
   5.1.3Experiments and Results
•  
  5.2Hand Based Robot Control
•  
   5.2.1Background
•  
   5.2.2System Description
•  
   5.2.3Algorithms Overview and Results
•  
   5.3Summary
•  
  Discussion and Conclusion
•  
   6.1Conclusions
•  
   6.2Limitations
•  
   6.3Suggestions for Future Works
•  
   Appendix A - Expectation Maximization
•  
   Appendix B- The CONDENSATION Algorithm
•  
   Appendix C- The MultiCam's Data Sheet
•  
   Bibliography