IOT-Based Physical Contact Tracing and Distance Applications To Prevent Covid-19 Transmission

Abstract

Social Distancing is an act of preventing viral infections such as COVID19 by encouraging healthy people to limit direct contact with other people. The implementation of Social Distancing includes not being allowed to shake hands, and maintaining a distance of at least 1 meter when interacting with other people. This research aims to build a system based on deep learning technology by utilizing cameras that can monitor simultaneously tracing Physical Contact or Social Distancing Violations that occur in a closed room remotely online via the web or locally connected to a web-based application. The tools used are Raspberry Pi devices as microcomputers and several supporting tools such as webcams, speakers, and web servers. This system utilizes the YOLO (You Only Look Once) Algorithm that has been trained on the COCO data set to identify human objects in image frames, then uses a deep learning algorithm with the OpenCV library to estimate the distance between humans in frames. The system must be calibrated with respect to real distance (meters) and image distance (pixels) according to the viewing angle of the camera installation. By implementing two algorithms, the system reports the number of violators, screenshots, and time when a social distancing violation occurs on a website-based application based on the set threshold distance (pixel). At the same time the Speaker on the device sounds to warn that a Social Distancing violation has occurred. Several tests were carried out in this study, namely testing how accurate the YoLo algorithm and object distance calculations that had been made were in monitoring an office room activity, measuring the comparison of real distance (meters) and frame distance (pixels), also observing how long it took the Raspberry Pi to send data. Social Distancing violations to Database. A comparison test of the YoLov3 and YoLov3-tiny algorithms was also carried out in processing the Social Distancing program in terms of Realtime performance.