A CNN-IOT Model for Suspicious Behaviour Detection of People by Surveillance using ML and DL

This project aims to create an intelligent surveillance system with IoT integration that automates patrol functions for improved safety and security protection for sensitive areas as buses, airports, banks, schools. Using deep learning algorithms, the system examines surveillance cameras for any unusual actions in real-time. The suggested method combines object detection with YOLOv8, image processing using OpenCV, and activity recognition via CNN-based classifiers. These models are jointly realized with body pose estimation, human action tracking, and analysis of behavior in real time. In the case where the actions of a person are suspicious, like loitering, fighting, or unauthorized entry to a restricted area, an alarm is instantaneously activated. It will replace one or multiple cameras with a programmable camera that can be repositioned or focused remotely for real-time surveillance of IoT-enabled systems, making it possible to monitor from afar. The proposed system is part of the IoT ecosystem, which removes the issue of monitoring remotely in real time. The system works in bad lighting and crowded conditions, minimizing the chances of a lot of false alarms and avoiding human error. The system's efficiency and flexibility of being deployed in all types of surveillance environments is emphasized with receiving over 92 % accuracy in identifying suspicious behavior as the main experimental result. The use of deep learning and automation will serve as a more effective method for perimeter control.