AI-powered CCTV for intuder and visitor detection with IoT alert system

Fong, Zhan Yet (2025) AI-powered CCTV for intuder and visitor detection with IoT alert system. Final Year Project, UTAR.

PDF Download (5Mb)

Abstract

AI-powered technologies that offer real-time detection and response are slowly replacing passive CCTV in surveillance systems. Traditional systems require constant monitoring and lack instant alert mechanisms, creating gaps in accessibility and responsiveness. This project addresses these limitations by developing a smart CCTV system that integrates artificial intelligence with IoT for improved security monitoring. The main objective is to design and implement an AI-powered CCTV capable of detecting intruders and visitors in real-time while providing instant mobile alerts and door control. The system was implemented using Python, OpenCV, Dlib for face recognition, YOLOv8n for human detection, and Telegram Bot API for IoT-based alerts. An Arduino-controlled servo motor was added to simulate automated door access for authorized users. Motion detection with MOG2 background subtraction triggered further analysis to balance efficiency and accuracy. Experimental results showed reliable performance across modules. Motion detection achieved high accuracy under normal conditions, face recognition reached up to 90% accuracy, and YOLOv8n human detection recorded an F1 score of 0.922 on test data. Telegram alerts were delivered within 1–2 seconds, ensuring timely notifications. In conclusion, the developed prototype demonstrates that integrating AI and IoT can deliver an affordable and effective smart surveillance system. While performance drops in lowlight conditions remain a limitation, the system provides a practical foundation for scalable home security with potential for future enhancements such as night vision integration, latency reduction, and expanded IoT features.

Actions (login required)