Real-Time Weapon Detection and Restricted Zone Intrusion Monitoring Using Deep Learning and Spatial Analysis

Abstract

Security surveillance systems play a vital role in safeguarding public and private spaces. However, traditional surveillance systems rely heavily on manual monitoring, which is in- efficient and prone to delayed responses. This research presents a real-time surveillance alert system capable of detecting knives in live video streams and identifying intrusions within user-defined restricted zones. The system integrates a YOLO-based deep learning model for weapon detection with polygon-based spatial reasoning for intrusion analysis. The proposed framework facilitates accurate threat identification, real-time alert genera- tion, and visual annotation of security breaches.