Weapons Identification Based Violence Detection in Real-time Surveillance System for Public Safety

Abstract

In a world where ensuring public safety is of paramount importance, the use of surveillance cameras for violence detection has gained significant attention. One critical aspect of this is the detection and prevention of violent activities in various settings, such as public spaces, schools, and commercial establishments. Surveillance cameras play a pivotal role in monitoring these environments, but the sheer volume of footage generated can overwhelm human operators. To address this challenge, here propose a novel approach to violence activity classification using weapons detection through surveillance camera systems. This proposed work presents an innovative approach that leverages the YOLO (You Only Look Once) object detection method to address this imperative. Modern surveillance systems generate vast volumes of data, necessitating efficient methods to monitor and respond to potential threats. To confront this challenge, this method capitalizes on cutting-edge computer vision techniques, specifically employing the YOLO algorithm to automatically analyze real-time video streams from surveillance cameras. Beyond object detection, it also designed to interpret the context of the detected weapons. It analyzes the temporal aspects of video sequences, distinguishing between benign handling and threatening behavior involving the detected weapons. When the system identifies potential violence, it generates real-time alerts to notify security personnel or relevant authorities. These alerts can include visual cues highlighting the suspicious activity in the surveillance feed, enhancing response time and accuracy.