Login Search
FMDB Transactions on Sustainable Critical Infrastructures Submit Paper
    Crossref logo

    Archiving

    Portico

    Fernando Martins De Bulhão is a member of

    Crossref logo

An Edge-Based Deep Learning Surveillance System Using YOLOv8 for Real-Time Multi-Zone Intrusion Detection and Intelligent Alerting

Authors:
Yashmit Rai, Girik Sharma, V. Kamalesh, R. Regin, Maruf Farhan

Addresses:
Department of Artificial Intelligence and Machine Learning, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of STEM, University of Sussex International Study Centre, Brighton, England, United Kingdom.

Abstract:

The need for reliable and independent surveillance systems has witnessed an increase in the recent past, especially in security-critical areas such as restricted areas, industrial areas, and learning centres. However, conventional detection-based surveillance systems are known to trigger alarms for all detected targets, regardless of their spatial importance, leading to numerous false alarms and undermining the effectiveness of the system. To mitigate the drawbacks of conventional detection-based surveillance systems, this paper proposes an innovative approach to develop a surveillance system using an edge-based intrusion detection system in conjunction with a hierarchical multi-zone spatial decision algorithm. In this proposed system, the entire area of interest is divided into three zones: Safe, Warning, and Intrusion. Alarms are generated only for targets detected in the Intrusion zone of the surveillance area. The system is designed to be independent and self-sufficient, without relying on cloud infrastructure, ensuring immediate responses and maintaining the privacy of surveillance data. Further, the system has utilised Telegram for real-time alerts and the web-based system for real-time surveillance. The results show that false-positive alerts are reduced to a minimum with spatially aware object detection, enabling efficient real-time performance and making the system suitable for real-time scenarios.

Keywords: Edge Computing; Intrusion Detection; Multi-zone Classification; Computer Vision; Surveillance Systems; Real-time Surveillance; False Positive Reduction; Independent Surveillance.

Received on: 01/08/2025, Revised on: 20/09/2025, Accepted on: 28/11/2025, Published on: 31/03/2026

DOI: 10.69888/FTSCIS.2026.000710

FMDB Transactions on Sustainable Critical Infrastructures, 2026 Vol. 1 No. 1, Pages: 16-32

  • Views : 31
  • Downloads : 11
Download PDF