Representation and Recognition of Agent Interactions Using Marking Analysis in Generalized Stochastic Petri Nets

This paper presents a novel approach for video event representation and recognition of multi agent interactions. The proposed approach integrates behavior modeling techniques based on Generalized Stochastic Petri Nets (GSPN) and introduces Petri net marking analysis for better scene understanding. The GSPN model provides remarkable flexibility in representation of time dependent activities which usually coexist with logical, spatial and temporal relations in real life scenes. The nature of Petri net concept allows efficient modeling of the complex sequential and simultaneous activities but disregards the global scope of a given model. The proposed marking analysis creates a new model extension based on the global scene view and uses historical and training information for current and future state interpretations. The GSPN approach is evaluated using the developed surveillance system which can recognize events from videos and give a textual expression for the detected behavior. The experimental results illustrate the ability of the system to create complex spatio-temporal and logical relations and to recognize the interactions of multiple objects in various video scenes using GSPN and marking analysis capabilities.