Security surveillance robot using machine learning techniques

Abstract

The aim of this project is to design a security surveillance robot target at general home users. Demand for surveillance using edge devices has surged due to end user’s need for security system to deliver convenience, cost effectiveness and privacy protection. This project loads Raspberry Pi 3 Model B with You Only Look Once (YOLO) model to allow speedy real-time image processing. The CNN model provides a 45 fps processing rate which is close to real time. A robot has been built to carry the Raspberry Pi 3. This would empower the powerful microcomputer with swift mobility. This Security Surveillance Robot can roam about in user household to monitor behaviours and activities. It detects events and recognize pre-trained behaviours. Upon analysis of the recognized anomalies from normal living, end users will be notified by mobile device. The light weight model loaded onto the robot eliminate the need to upload user’s data to the service providers. Data capture as well as analysis will all be done on the robot, locally.