Virtual Fencing for Safety-Critical Cyber-Physical Systems: Computer-Vision Enabled Digital Twins

Abstract

Early warning zones (EWZs) are pivotal for future crowd management in smart cities, leveraging computer vision to transform dynamic environments into controllable cyber-physical systems. This approach aims to restrict unsafe or threatening movements by creating EWZs that enhance the resilience of critical infrastructures and ensure citizen safety. While conventional virtual fencing uses GPS-based solutions for outdoor zone-monitoring or surveillance of densely populated areas, as well as fixed cameras for indoor environments (e.g., museums), this article explores the use of computer vision and Uncrewed Aerial Vehicles (UAVs) to create risk-ranked EWZs. These zones can safeguard critical infrastructure and densely populated areas by using UAVs (where fixed cameras are not installed), thus promising to enhance crowd management and safety in smart cities. In this study, the EWZs are categorized by risk levels, with the proximity to hazardous areas determining the severity from low to high. This tiered structure allows for appropriate and timely responses to potential threats, thereby ensuring a robust early warning mechanism. A physical testbed was constructed to monitor human movement as a reflection of behavior within this cyber-physical-social environment. Experiments simulating virtual fence (V-fence) crossings demonstrated the system’s effectiveness in providing early warnings. The results also showed that the system successfully tracked multiple persons through a lightweight framework that can be deployed at the edge, ensuring real-time surveillance and response.