Child Safety Monitor in Cars Using Machine Learning

This Child Safety Monitoring System transmits a warning to the parent or guardian if a child is left inadvertently in the car. The deaths related to children left in cars recur even though many systems thus far exist. This developed system aims at making a move in preventing these accidents. This system can classify images in real-time using Machine Learning, which consists of an Arduino Pro Mini board equipped with an IR Proximity Sensor to detect whether the car doors are closed or not. An own, unique dataset has been developed with eight individual children. The Haar-like feature algorithm is chosen to detect individual faces, which uses Haar features and Integral Images and helps at the stage of feature extraction. The Local Binary Pattern Histogram has been utilized for facial recognition. The suggested study shows that the LBPH algorithm is a more popular and accurate face recognition algorithm when compared to the Eigenfaces and Fisherfaces algorithm. This LBPH algorithm treats the result as a binary number and visualizes the final result as histograms. The detected child face is sent to the parent or guardian through e-mail only in case of the doors and windows being closed. The results imply that the LBPH algorithm approach generates promising performances for the task of detecting child-like objects, with the highest accuracy of 93.4% and an average accuracy of 89.27%.