Authors:
Selim Benhimane
1
;
Hesam Najafi
2
;
Matthias Grundmann
3
;
Yakup Genc
2
;
Nassir Navab
1
and
Ezio Malis
4
Affiliations:
1
Technical University Munich, Germany
;
2
Siemens Corporate Research, Inc., United States
;
3
College of Computing, Georgia Institute of Technology, United States
;
4
I.N.R.I.A. Sophia-Antipolis, France
Keyword(s):
Real-time Vision, Template-based Tracking, Object Recognition, Object Detection and Pose Estimation, Augmented Reality.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Computer Vision, Visualization and Computer Graphics
;
Feature Extraction
;
Features Extraction
;
Human-Computer Interaction
;
Image and Video Analysis
;
Image Formation and Preprocessing
;
Informatics in Control, Automation and Robotics
;
Matching Correspondence and Flow
;
Methodologies and Methods
;
Motion and Tracking
;
Motion, Tracking and Stereo Vision
;
Multi-View Geometry
;
Pattern Recognition
;
Physiological Computing Systems
;
Real-Time Vision
;
Signal Processing, Sensors, Systems Modeling and Control
Abstract:
Real-time tracking of complex 3D objects has been shown to be a challenging task for industrial applications where robustness, accuracy and run-time performance are of critical importance. This paper presents a fully automated object tracking system which is capable of overcoming some of the problems faced in industrial environments. This is achieved by combining a real-time tracking system with a fast object detection system for automatic initialization and re-initialization at run-time. This ensures robustness of object detection, and at the same time accuracy and speed of recursive tracking. For the initialization we build a compact representation of the object of interest using statistical learning techniques during an off-line learning phase, in order to achieve speed and reliability at run-time by imposing geometric and photometric consistency constraints. The proposed tracking system is based on a novel template management algorithm which is incorporated into the ESM algorithm
. Experimental results demonstrate the robustness and high precision of tracking of complex industrial machines with poor textures under severe illumination conditions.
(More)