In most recent Intelligent Video Surveillance systems, mechanisms to support human decisions are integrated in cognitive artificial processes. These algorithms mainly address the problem of extraction and modelling of relevant information from a sensor network. In crowd monitoring the main problem is to individuate specific events as for example different behaviours among interacting entities. A bio-inspired structure for modelling cause-effect relationships between events was lately proposed by the authors and applied to the field of automatic crowd monitoring. Such cause-effect relationships are modelled by means of coupled Event-based Dynamic Bayesian Networks and stored within an Autobiographical Memory during a learning phase, in order to supply appropriate knowledge to the automatic system in the on-line phase. However, the definition of causality relies on the selection of relevant events, which is performed by means of Self Organizing Maps and on a temporal scale defined by a newly introduced temporal parameter. Performances of the proposed multi-camera video surveillance system are studied on tuning such causality parameters.
Event definition for stability preservation in bio-inspired cognitive crowd monitoring
CHIAPPINO, SIMONE;MORERIO, PIETRO;MARCENARO, LUCIO;REGAZZONI, CARLO
2013-01-01
Abstract
In most recent Intelligent Video Surveillance systems, mechanisms to support human decisions are integrated in cognitive artificial processes. These algorithms mainly address the problem of extraction and modelling of relevant information from a sensor network. In crowd monitoring the main problem is to individuate specific events as for example different behaviours among interacting entities. A bio-inspired structure for modelling cause-effect relationships between events was lately proposed by the authors and applied to the field of automatic crowd monitoring. Such cause-effect relationships are modelled by means of coupled Event-based Dynamic Bayesian Networks and stored within an Autobiographical Memory during a learning phase, in order to supply appropriate knowledge to the automatic system in the on-line phase. However, the definition of causality relies on the selection of relevant events, which is performed by means of Self Organizing Maps and on a temporal scale defined by a newly introduced temporal parameter. Performances of the proposed multi-camera video surveillance system are studied on tuning such causality parameters.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
