A Vision-Based Social Distancing and Critical Density Detection System for COVID-19

Area of Science:

• Computer Vision
• Public Health
• Artificial Intelligence

Background:

• Social distancing (SD) is crucial for mitigating the spread of infectious diseases like Coronavirus Disease 2019 (COVID-19).
• Unintentional violations of SD measures can occur due to a lack of spatial awareness in crowded environments.
• Existing surveillance methods may raise privacy concerns or require human supervision.

Purpose of the Study:

• To propose an active, vision-based surveillance system for real-time detection and prevention of social distancing violations.
• To introduce a novel critical social density metric to maintain pedestrian density below a threshold.
• To ensure an ethically sound system that respects privacy and avoids individual targeting.

Main Methods:

• Development of a real-time, vision-based system utilizing deep learning models for detecting social distancing breaches.
• Implementation of non-intrusive audio-visual cues to alert individuals about potential violations.
• Definition and application of a critical social density value to manage pedestrian proximity.

Main Results:

• The system effectively detects social distancing violations in real-time.
• Maintaining pedestrian density below the critical social density value significantly reduces the occurrence of violations.
• The system operates ethically, without data recording, individual targeting, or human supervision.

Conclusions:

• The proposed active surveillance system is an effective tool for enforcing social distancing and slowing disease transmission.
• The critical social density metric provides a quantifiable approach to managing public spaces for safety.
• This AI-powered, privacy-preserving system offers a scalable solution for public health monitoring.