Survey of Social Distancing Detection and Effective Crowd Control Using Computer Vision for any Communicable Diseases

Abstract: Social distancing is still one of the most effective ways to control the spread of infectious diseases. This paper proposes a novel real-time social distancing monitoring and enforcement algorithm in public areas. In the background, we designed a novel algorithm that can automatically examine and quantify compliance to recommended social distancing practices using real-time video streams from cameras deployed on location. It uses sophisticated computer vision to complete a millisecond-by-millisecond spatial analysis of people. Our system accepts both passive monitoring and smart tracking. The solution relies on a tracking mechanism that detects people coming too close to each other against the prescribed distancing norms to identify deviations and immediately alert and guide them (Mobile App Users) into specific “Red Zones” when necessary. The research offers a major advancement in controlling diseases, providing an active real-time solution to enforcing social distancing measures in public places.

Keywords: Deep Learning, YOLOv5 / YOLOv6 / YOLOv7 algorithms, person detection, tracking, social distancing surveillance, Public Data Analysis, Computer Vision.

 Abbreviation:  YOLO (You Only Look Once), CNN (Convolutional Neural Network), COCO (Common Objects in Context), Single Shot Multibox Detector (SSD), ResNet (Residual Network), VGG (Visual Geometry Group).