Suspicious Activity Detection Using Live Video Analysis

Abstract

Video analytics is a field of computer science which involves processing of videos and live video streams to gather data for obtaining domain-specific information. The traditional closed-circuit television (CCTV) system requires humans to monitor the CCTV camera streams for 24/7 which is inefficient and costly. Additionally, the existing systems work in a post-offense mechanism, that is the video streams are checked only after a crime has already taken place. Expert video-surveillance tools can be used to automate this process and provide real-time solutions for different risk situations and helping the police officers or the intended group of people. This paper proposes an intelligent system, which utilizes an expert video-surveillance tool for monitoring of potentially suspicious and criminal activities in shopping malls. The system uses methods such as object detection, movement tracking and activity monitoring to robustly and efficiently track the location of objects and the subjects to identify questionable actions and activities. Along with this, an interface is developed to notify the concerned authority in real time. Thus, the system works as an assistant to the security personnel. The system uses various custom objects and person detection models to identify and establish a relationship between them.