Pose Estimation and Threat Identification : A Computer Vision Approach For Enhanced Surveillance Systems

Document Type : Research article

Authors

Mathematics and Computer Science Department, Faculty of Science, Alexandria University, Alexandria, Egypt.

Abstract

Abstract— In today’s security landscape, traditional surveil-
lance systems struggle with inefficiencies such as operator
fatigue, missed threats, and delayed responses. To address these
challenges, we propose a Smart Policing Surveillance System
leveraging advanced computer vision and deep learning. Our
system integrates human pose estimation, gesture analysis, and
weapon detection to identify both visible and concealed threats.
By analyzing body language and movements, it detects suspi-
cious behavior such as the positioning of arms and identifies
victims in distress through gestures like raised hands.
At its core, the system employs pose estimation for precise
body tracking and deep learning models for real-time threat
detection. Focusing on critical areas like hand positions, it re-
duces computational overhead while maintaining high accuracy.
Designed to perform reliably in challenging conditions such as
crowded environments, it triggers real-time alerts upon threat
detection, enabling swift responses from law enforcement.
Our experimental results demonstrate the system’s effective-
ness: it achieves 94.3% accuracy in pose estimation, 97.7%
weapon detection rate, and identifies raised hands with 89.8%
precision. The integrated solution detects more threats than
conventional systems while reducing false alarms . In field tests,
the system demonstrated faster threat identification compared
to human operators, proving its value as a vital tool for modern
security challenges.

Keywords

Main Subjects

Files

Share

How to cite

Statistics