Techniques for Parallel Execution of the Particle Filter

  • Ole-Christoffer Granmo
  • Frank Eliassen
  • Olav Lysne
  • Viktor S. Wold Eide
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2749)


Dynamic Bayesian networks are a promising approach to automatic video content analysis which allows statistical inference and learning to be combined with domain knowledge. When the particle filter (PF) is used for approximate inference, video data can often be classified in real-time (supporting e.g. on-line automatic video surveillance). Unfortunately, the limited processing resources available on a typical host restricts the complexity, accuracy, and frame rate of PF classification tasks. Here, we target this limitation by applying the traditional parallel pooled classifiers architecture to execute multiple PFs in parallel and to coordinate their output.We then identify a significant weakness of this approach in terms of loss of accuracy. To reduce the loss of accuracy, we propose a novel scheme for coordinating the pooled PFs based on the exchange of so-called particles. In an object tracking experiment, a significant loss of accuracy is observed for the naive application of the pooled classifiers architecture. No loss of accuracy is detected when our scheme for exchanging particles is used.


Particle Filter Directed Acyclic Graph Time Slice Object Tracking Parallel Execution 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    Jensen, F.V.: Bayesian Networks and Decision Graphs. Series for Statistics for Engineering and Information Science. Springer Verlag (2001)Google Scholar
  2. 2.
    Chang, S.F., Sundaram, H.: Structural and Semantic Analysis of Video. In: Multimedia and Expo 2000 IEEE. Volume 2. (2000) 687–690Google Scholar
  3. 3.
    Naphade, M., Huang, T.: Extracting semantics from audio-visual content: the final frontier in multimedia retrieval. IEEE Transactions on Neural Networks (2002) 793–810Google Scholar
  4. 4.
    Liu, J.S., Chen, R.: Sequential Monte Carlo methods for Dynamic Systems. Journal of the American Statistical Association 93 (1998) 1032–1044zbMATHCrossRefMathSciNetGoogle Scholar
  5. 5.
    Duda, R.O., Hart, P.E., Stork, D.G.: Pattern Classification. Wiley-Interscience. John Wiley and Sons, Inc. (2000)Google Scholar
  6. 6.
    Eide, V.S.W., Eliassen, F., Granmo, O.C., Lysne, O.: Scalable Independent Multi-level Distribution in Multimedia Content Analysis. In: Protocols and Systems for Interactive Distributed Multimedia (IDMS/PROMS2002). Lecture Notes in Computer Science 2515, Springer (2002) 37–48CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2003

Authors and Affiliations

  • Ole-Christoffer Granmo
    • 1
  • Frank Eliassen
    • 2
  • Olav Lysne
    • 2
  • Viktor S. Wold Eide
    • 2
  1. 1.Department of Information and Communication TechnologyAgder University CollegeGrimstadNorway
  2. 2.Simula Research LaboratoryLysakerNorway

Personalised recommendations