Artificial vision algorithm in Raspberry, to remotely monitor an infant while he remains in his crib

Authors

Vidal, Cristopher

Huamaní, Pedro

Tinoco Plasencia, Christian

Puelles Bulnes, Elizabeth

Abstract

This article shows the algorithmic implementation of artificial vision techniques in a Raspberry Pi 3 Model B+ microcomputer, for monitoring during the day and real time of an infant who was represented by a doll, and while he remains in his cradle; so that, direct supervision is maintained, providing peace of mind and security to parents. For it, it was decided to serially integrate an ATMEGA 328p controller to control the actuators and acquire the variable weight of the infant, an Arduino module for connection with a distraction toy located in the cradle itself, a Shield Sim900 GSM module for sending text messages in order to alert about the current situation of the infant, and a Raspberry microcomputer for the image processing stage, which were acquired through a video camera in charge of focusing and monitoring the infant's cradle, and to know the position of the infant and thus carry out an action regarding each particular detection case (position ulna dorsal and ventral, side and fallen). Secondly, in the image processing stage, the Template Matching technique was used to find areas like a template, and was complemented with the Normalized Difference function, which delivered better detection results in terms of five other functions.