Current methods for detection of crowd behavior are based on tracking individuals or virtual particles. Such assessments are highly prohibitive, preventing the real time and efficient analysis. Other approaches use comparisons of individual behaviors with normal patterns. However, these approaches are tedious and unreliable. This novel procedure allows for the use of a sensing system (comprising video and multi-sensor information) for real-time detection of the behavior of a large, dense crowd.
Value Proposition
The procedure/system:
•Allows for a layered, multi-scale allocation of sensors and computational resources to address prohibitive complexities as associated with conventional procedures
•Enables a high resolution, fine grain analysis of small areas associated with a large crowd
•Identifies any anomaly as a discrepancy between a normal and an observed behavior
•Renders the anomaly detection problem as a simple dynamic estimation/assessment of few model parameters
•Would be commercially useful for the following applications
o Large crowd management in both normal and emergency conditions
o Detection of suspicious behaviors as in terrorist activities
o Management of large sports arenas, pedestrian malls, political rallies, large occupancy buildings, transportation terminals, etc.
Intellectual Property status
PCT Application PCT/US2012/070149
License status
Available for license
Direct Link:http://neu.technologypublisher.com/technology/12111