Bayesian network based multi stream fusion for automated online video surveillance

Dejaii Arsić, Frank Wallhoff, Björn Sdvuller, Gerhard Rigoll

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

Video Surveillance is an omnipresent topic when it comes to enhancing security in public places and transportation systems. Fully automated behavior detection systems are desirable when it comes to cutting costs for analysing video and audio streams online. These will initiate an alarm signal autonomously if a possibly dangerous situation is detected. The particular investigated scenario is monitoring passengers' behaviors in aircrafts. In order to work robustly in unconstrained environments many subsystems have to be developed Though in the last years reliable approaches far required systems have been brought up, there exists a gap between reliability and computational effort. Hence a Law Level Activity representation of behaviors will be presented, which can be detected with so called weak classifiers in real time. These output will be interpreted by a highly sophisticated probabilistic Bayesian Network.

Original languageEnglish
Title of host publicationEUROCON 2005 - The International Conference on Computer as a Tool
PublisherIEEE Computer Society
Pages995-998
Number of pages4
ISBN (Print)142440049X, 9781424400492
DOIs
StatePublished - 2005
EventEUROCON 2005 - The International Conference on Computer as a Tool - Belgrade
Duration: 21 Nov 200524 Nov 2005

Publication series

NameEUROCON 2005 - The International Conference on Computer as a Tool
VolumeII

Conference

ConferenceEUROCON 2005 - The International Conference on Computer as a Tool
CityBelgrade
Period21/11/0524/11/05

Keywords

  • Bayesian networks
  • Low level features
  • Multi stream fusion
  • Video surveillance

Fingerprint

Dive into the research topics of 'Bayesian network based multi stream fusion for automated online video surveillance'. Together they form a unique fingerprint.

Cite this