Multimodal Video Summarization For Crime Scene Analysis Using Yolo Based Object Detection

Abstract

The Rapid Growth Of Surveillance Cameras In Public Places Has Led To The Generation Of Large Volumes Of Video Data, Making Manual Monitoring Difficult And Inefficient. Detecting Violent Activities Such As Fights, Weapon Usage, And Suspicious Behaviour In Real Time Is Essential For Maintaining Public Safety. This Paper Proposes A Deep Learning-based System For Crime Scene Analysis Using Multimodal Video Summarization And YOLO-based Object Detection. The System Processes Surveillance Videos By Extracting Frames And Applying Preprocessing Techniques, Followed By Object Detection Using The YOLO Algorithm. Convolutional Neural Networks (CNN) Are Used For Feature Extraction To Identify Objects Such As People And Weapons. The System Analyses Detected Objects To Identify Suspicious Activities And Generates Real-time Alerts With Relevant Information Such As Timestamps And Confidence Scores. This Approach Improves The Efficiency Of Surveillance Monitoring And Helps Security Personnel Respond Quickly To Potential Threats.

Keywords