The program developed a wide range of video capabilities relevant to law enforcement and corrections scenarios. The technology developed can assist appropriately trained law enforcement and correctional personnel to recognize suspicious behavior that may precede disorderly conduct and criminal behavior, including prison fights and riots, the formation of drug markets in a community, and gang activities. The program produced five major achievements. First, a resource description framework (RDF) was created to dynamically represent and maintain probabilistic and non-probabilistic data, which are the basis for trained personnel to recognize probabilistic events. Second, a probabilistic event-recognition system was developed that combines low-level probabilistic evidence and rule-based domain knowledge that enables the detection of pre-defined events from video images of past or real-time events. Third, features such as event-explanation, and scenario-modeling GUI were implemented in order to increase the system's relevance to law enforcement and corrections work. Fourth, the project developed a novel framework for learning-based event recognition that can achieve reliable analysis of real-time video transmissions. Fifth, the system was tested live during the 2010 Mock Prison Riot sponsored by the U.S. Justice Department's National Institute of justice, as well as real-world video data collected from the surveillance camera network at Schenectady, NY. 107 figures, 11 tables, and 126 references
Advanced Behavior Recognition in Crowded Environments
NCJ Number
240575
Date Published
September 2011
Length
217 pages
Annotation
Features and outcomes are described for a program that increases situational awareness by law enforcement and correctional personnel, which led to the reliable detection and prevention of disorderly conduct and criminal behavior.
Abstract
Date Published: September 1, 2011