A Practical Approach to Real-Time Intelligent Surveillance Using YOLOv8 and DeepSORT with Behavior Analysis and Zone-Based Monitoring

The proposed AI-Powered Intelligent Surveillance System successfully demonstrates a robust and scalable solution for real-time security monitoring using advanced deep learning and computer vision techniques. The system effectively integrates human detection using YOLOv8, multi-object tracking using DeepSORT, behavior analysis, restricted zone monitoring, and real-time alert generation into a unified framework. The experimental results confirm that the system is capable of accurately detecting and tracking multiple individuals while maintaining consistent identity across frames, even in dynamic and crowded environments. The inclusion of behavior analysis significantly enhances the system’s capability by enabling real-time classification of human activities such as walking, sitting, falling, and fighting. The system also effectively identifies abnormal behaviors and highlights them with confidence scores, improving situational awareness and enabling faster response to critical events. The polygon-based restricted zone monitoring further strengthens the system by detecting unauthorized entry with high precision and triggering instant alerts with supporting evidence. The alerting and event logging mechanisms ensure that all critical events are recorded with detailed metadata, including timestamps, tracking IDs, detected behaviors, and snapshot evidence. The dashboard interface provides an intuitive and real-time visualization of surveillance activities, allowing security personnel to monitor, analyze, and respond efficiently. Overall, the system achieves high accuracy, low latency, and reliable performance across different conditions, reducing the dependency on manual monitoring and minimizing the chances of missed incidents. The modular architecture ensures scalability and adaptability, making the system suitable for deployment in real-world environments such as campuses, offices, public spaces, and high-security zones. The results demonstrate that the proposed system provides an effective, intelligent, and automated surveillance solution capable of enhancing security and operational efficiency.

[1] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards Real-Time Object Detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137–1149, 2015. [2] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You Only Look Once: Unified, Real-Time Object Detection,” arXiv preprint, 2016. [3] N. Wojke and A. Bewley, “DeepSORT: Simple Online and Real-Time Tracking,” arXiv preprint, 2018. [4] X. Li, R. Zhao, and L. Li, “Human Tracking Using SORT Algorithm,” International Journal of Computer Vision, vol. 127, pp. 1499–1513, 2019. [5] A. Bochkovskiy, C. Wang, and H. Liao, “YOLOv4: Optimal Speed and Accuracy,” arXiv preprint, 2020. [6] R. Patel, A. Kumar, and S. Singh, “Smart Video Surveillance Using Computer Vision,” Journal of Ambient Intelligence and Humanized Computing, vol. 11, pp. 1231–1244, 2020. [7] C. Zhang, Y. Li, and H. Wang, “AI-Based Surveillance in Public Safety,” IEEE Access, vol. 9, pp. 55612–55623, 2021. [8] M. Karthik, K. Ramesh, and V. Suresh, “Deep Learning for Anomaly Detection,” Computers & Security, vol. 103, 2021. [9] A. Singh, R. Sharma, and P. Verma, “Intrusion Detection Using Deep Learning in CCTV,” Journal of Intelligent & Fuzzy Systems, vol. 42, no. 3, pp. 3151–3162, 2022. [10] P. Sundaramoorthy, “AI in Intelligent Security Systems,” International Journal of Security and Networks, vol. 17, no. 2, pp. 101–114, 2022. [11] G. Jocher et al., “YOLOv8: Ultralytics,” arXiv preprint, 2023. [12] Y. Chen, F. Li, X. Zhao, and Q. Wang, “Vision-Based Real-Time Alert Systems,” IEEE Transactions on Industrial Informatics, vol. 19, no. 4, pp. 2345–2356, 2023.