Autonomous Human Violence Detection In Smart Surveillance

Abstract

Autonomous Human Violence Detection Has Become An Essential Component In Modern Smart Surveillance Systems Due To The Increasing Demand For Public Safety And Automated Monitoring. Traditional Surveillance Methods Rely Heavily On Continuous Human Observation, Which Is Inefficient And Prone To Errors When Monitoring Multiple Video Streams For Long Durations. To Address This Limitation, This Paper Presents An Autonomous Human Violence Detection System For Smart Surveillance Using Deep Learning And Real-time Video Analysis. The Proposed System Uses A Laptop Camera To Capture Live Video And Analyzes Human Activities Using A Trained Deep Learning Model Developed From Two Different Datasets Consisting Of Normal Human Actions And Violent Activities. Convolutional Neural Networks (CNN) And Object Detection Techniques Are Used To Extract Features From Video Frames And Classify The Actions As Normal Or Violent. The Trained Model Compares Live Video Input With Learned Patterns And Generates Real-time Predictions To Identify Suspicious Behavior. Experimental Results Show That The Proposed System Can Detect Violent Activities With Good Accuracy And Can Be Applied In Public Surveillance, Educational Institutions, And Security Monitoring Environments, Reducing The Need For Continuous Human Supervision And Improving The Efficiency Of Intelligent Surveillance Systems.

Keywords