Detection & Recognition of Undesired Events in Surveillance Videos

Abstract

CCTV (Closed Circuit Television System) surveillance and security has garnered its importance throughout the years. There has been a significant increase in the installation of surveillance cameras in universities, schools, retail stores, banks, shopping centers, airports, secluded areas such as parking lots and back alleys. Surveillance helps to observe, monitor, control, detect, identify and recognize situations and individuals that are potentially harmful to other persons and property. Active shooter events and mass murder incidents show that the detection and recognition of such events beforehand leads to a rapid response and reduction in the number of casualties. The need of autonomously analyzing and identifying security threats becomes massively important. The aim of this project is to develop a system that can analyze videos and perform the task of detection and recognition of undesired events. Firearms held by a person in their hand are one of the most common items used in assaults. They are a primary sign of danger. This system will be capable of generating early warnings to help prevent any undesired event take place. This will result in reducing the number of potential victims and lead to an effective and faster response. This project also aims to detect fighting events as they are a clear sign of threat. This is particularly useful in surveillance as it may be possible to intervene to stop a crime at the earliest possible opportunity. The project also focuses on the detection of person climbing a fence or a wall, which is also deemed to be an unwanted activity. All these activities cause an enormous load for the surveillance operators, as the number of camera scenes a single operator monitor’s is limited. The system will help in reducing this load.