Intelligent Video Monitoring System with the Functionality of Online Recognition of People’s Behavior and Interactions Between People

Abstract

The intelligent video monitoring system SAVA has been implemented as a prototype at the 9th Technology Readiness Level. The source of data are video cameras located in the public space that provide HD video streaming. The aim of the study is to present an overview of the SAVA system enabling identification and classification in the real time of such behaviors as: walking, running, sitting down, jumping, lying, getting up, bending, squatting, waving, and kicking. It also can identify interactions between persons, such as: greeting, passing, hugging, pushing, and fighting. The system has module-based architecture and is combined of the following modules: acquisition, compression, path detection, path analysis, motion description, action recognition. The effect of the modules operation is a recognized behavior or interaction. The system achieves a classification correctness level of 80% when there are more than ten classes.