Intelligent Social Distancing Enforcement: A YOLOv3-Based Framework for Health Security

Abstract

During the COVID-19 pandemic, individuals encountered numerous challenges due to the swift spread of the virus. Effective implementation of social distancing is crucial in halting the transmission of airborne, touch-based, and close-contact diseases such as Chickenpox, Influenza, Pertussis (whooping cough), and Respiratory Syncytial Virus (RSV), among others. Regrettably, many individuals are not adhering to these prescribed norms. This study aims to address the issue by introducing a system for close contact detection and assessing the proximity between individuals to mitigate the impact of contagious diseases. Our framework utilizes the YOLOv3 algorithm for human identification in video sequences, extracting information from detected bounding boxes. The pairwise distances are computed using the Euclidean distance formula. A tracking algorithm is then employed to monitor individuals throughout the video sequence. In gauging compliance with social distancing guidelines, we set a threshold for detecting violations and evaluating whether the distance between individuals falls below the minimum prescribed social distance. Additionally, a mailer function is being implemented to notify authorized personnel if any violations of social distancing norms occur, subject to certain predefined conditions being met.