BAT - A distributed meta-tracking system

In proceedings of Eurographics Symposium on Virtual Environments (EGVE), Mai 2008
 

Abstract

This paper describes the design of the 'BAT' (Bonn Articulated Tracker) visual tracking framework. This system allows the easy implementation of real-time, multi-camera motion tracking that can be distributed (also in multithreaded sense) across several computing nodes (or CPU cores). The system in itself does not realize any specific tracking system, but manages a meta-algorithm flow between processing blocks. An actual tracking implementation is realized by specifying the processing blocks through plugins. Depending on the plugins supplied, 'BAT' is capable to instantiate a wide-variety of systems ranging from object-detection methods to model-based deformable object tracking based on time-coherence, allowing also for hybrid algorithms. Being a "meta dataflow system", 'BAT' also naturally facilitates sensor fusion. Moreover, it can be used as a testbed to compare and evaluate different kind of tracking algorithms or algorithm substeps.

Paper herunterladen

Paper herunterladen

Bibtex

@INPROCEEDINGS{kahlesz-2008-bat,
     author = {Kahlesz, Ferenc and Klein, Reinhard},
      title = {BAT - A distributed meta-tracking system},
  booktitle = {Eurographics Symposium on Virtual Environments (EGVE)},
       year = {2008},
      month = may,
    address = {Eindhoven, The Netherlands},
   abstract = {This paper describes the design of the 'BAT' (Bonn Articulated Tracker) visual tracking framework.
               This system allows the easy implementation of real-time, multi-camera motion tracking that can be
               distributed (also in multithreaded sense) across several computing nodes (or CPU cores). The system
               in itself does not realize any specific tracking system, but manages a meta-algorithm flow between
               processing blocks. An actual tracking implementation is realized by specifying the processing blocks
               through plugins. Depending on the plugins supplied, 'BAT' is capable to instantiate a wide-variety
               of systems ranging from object-detection methods to model-based deformable object tracking based on
               time-coherence, allowing also for hybrid algorithms. Being a "meta dataflow system", 'BAT' also
               naturally facilitates sensor fusion. Moreover, it can be used as a testbed to compare and evaluate
               different kind of tracking algorithms or algorithm substeps.}
}