Event Template Hierarchies as Means for Human-Automation Collaboration in Security Surveillance

Abstract

Advances in remote sensing systems provide human monitors access to more data. The current challenge is to help extract relevant patterns and direct the attention of human monitors and human problem holders to the changing security picture with respect to acute situations (a normal market scene turns into an ethnic confrontation) or longer term trends (seeing new patterns of 'typical' behavior to avoid false alarms). Security surveillance monitoring can be advanced through new event recognition capability of autonomous monitors and by effectively coupling these sensor/algorithm systems to human monitors and problem holders. To meet these challenges in security surveillance, we have developed an event-sensitivea rchitecturew here machine agents provide event-basedi nformationt o human monitors and problem holders and are re-directable given contextual information. The key innovation is a context-based hierarchical event template structure which can be used to integrate data over a distributed sensor system.