de
en
Schliessen
Detailsuche
Bibliotheken
Projekt
Impressum
Datenschutz
zum Inhalt
Detailsuche
Schnellsuche:
OK
Ergebnisliste
Titel
Titel
Inhalt
Inhalt
Seite
Seite
Im Dokument suchen
Ghobadi, Seyed Eghbal: Real time object recognition and tracking using 2D/3D imagesObjekterkennung und -verfolgung in Echtzeit mit Hilfe von 2D/3D Bildern. 2010
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