AI-Powered Smart Surveillance Camera System

Abstract

The widespread deployment of surveillance cameras in both public and private domains has intensified the need for intelligent monitoring solutions that can surpass the limitations of human-based observation. Continuous manual supervision of multiple video streams often leads to fatigue, reduced attention, and delayed reactions, increasing the likelihood of overlooked critical events.To overcome these challenges, this paper presents an AI-driven smart surveillance system that integrates real-time video processing with deep learning techniques for automated monitoring. The proposed system utilizes advanced object detection models such as YOLOv5 and YOLOv8 to achieve fast and accurate identification of humans and relevant objects within video frames. In addition, centroid-based tracking is employed to maintain consistent identification of individuals across consecutive frames.The system is capable of recognizing potentially suspicious behaviors, including unauthorized intrusion, prolonged loitering, abnormal crowd gathering, and unattended objects. Detected events are systematically recorded using a lightweight SQLite database, ensuring efficient data storage and retrieval. A web-based interface developed using Flask enables live video streaming with detection annotations and real-time event updates for users.By reducing reliance on manual monitoring and incorporating intelligent filtering mechanisms, the system minimizes false alarms while improving overall surveillance effectiveness. Furthermore, its scalable architecture and compatibility with IoT frameworks make it suitable for deployment in smart cities, industrial environments, educational campuses, and high-security zones.