Home Intruder Detection System using Machine Learning and IoT

Abstract

Home surveillance requires human effort, time and cost. Many tragedies such as robbery and vandalism occurred at home while the owners were negligent or not at home. Some residential areas hire guards to monitor their homes but hiring workers is not considered a cost-efficient option. Home Intruder Detection System (HIDES) is an Internet of Things (IoT) system with a mobile application to help homeowners in house surveillance by alerting users for any potential threats remotely. The main objectives of HIDES are to create a reliable home security system with the implementation of IoT, to implement the object detection algorithm to determine the presence of humans, and to develop a smart mobile application for users to monitor their houses from anywhere in the world and be alerted if any threats are detected. HIDES is developed using the System Development Life Cycle (SDLC) approach. HIDES implements an object detection algorithm; Single-Shot Multibox Detection (SSD) in NVIDIA Jetson Nano to detect intruders through a camera connected to the system. HIDES successfully achieves its objective in detecting persons precisely and alerting the detection to users through mobile application remotely. The system can capture video at an average of 20 frames per second (FPS) while detecting intruders and sending detection video to the server. The mobile application achieves good performance where the loading time takes 2.3 seconds while only requiring about 0.99MB of memory to run and 66.87MB of space