Vehicle's Interior Presence Detection and Notification System

Abstract

This thesis aims to present both a solution analysis and potential implementation to prevent in-car infants' deaths, capitalizing on low-cost technologies that can easily be integrated on the vehicle or are already a part of it. The circumstances that lead to this outcome can be narrowed down to parent's carelessness, whether they left the child unattended because of a short time stop to do groceries that turned out too long, or simply because it was forgotten after bringing such groceries home first. As part of the state of art, several possible systems were considered, culminating in a system idealization incorporating two approaches as inputs - a motion sensor and a camera - while notifying the driver via SMS. In order to study the advantages a division was made, resulting in a low-cost approach, which utilizes the motion sensor and other that is anticipated to be more reliable consisting in human detection through vision algorithms. While the integrated system is idealized to optimally recognize the presence of children inside the vehicle, the performed studies premise a potential stand-alone application. Throughout this study it is noted the low-cost implementation limitations due to the sensor requiring movement to acknowledge that there is someone present, while the use of a camera can recognize infants whether or not moving. At the same time, vision algorithm improvement becomes even more reliable overtime, as well as the car electronics, with cameras being used for several other issues such as driver drowsiness detection, mitigating cost constraints. It is hoped that this thesis serves as both a solution and solution catalyst since it is not limited to this application, being able to prevent cars from being stolen or burglar's break-in to surprise an unware driver, although it's main focus is to present a reliable detection and notification system that allows in-car infant casualties decrease, turning such deaths into deliberate crimes.