Smart Surveillance System for Intrusion Detection and Behaviour Analysis

Abstract: This project designs and implements a Smart Surveillance System that integrates artificial intelligence and Internet of Things (IoT) technology to enable real-time intrusion detection and human behavior analysis. The system utilizes ESP32-CAM modules equipped with infrared night vision and PIR motion sensors to continuously monitor both indoor and outdoor environments. By employing state-of-the-art deep learning models such as YOLOv8 and MobileNet, it accurately detects human presence and identifies unauthorized entries. Furthermore, the system uses MediaPipe for advanced pose estimation to recognize and classify specific behavioral patterns—including falling, sudden running, prolonged loitering, and trespassing. Upon detecting any anomalous activity, the system triggers multi-channel alerts including a audible buzzer, SMS notifications to authorized personnel, and detailed email reports with incident snapshots. A centralized web dashboard built with Flask provides real-time video streaming, interactive event logs, and remote configuration capabilities. By storing only relevant event-triggered footage, the system significantly optimizes storage usage and reduces costs. Scalable and hardware-agnostic, this solution is suitable for diverse applications such as educational laboratories, elderly care facilities, industrial sites, and smart homes, offering a proactive, cost-effective, and intelligent alternative to traditional surveillance systems.