Using Computer Vision to Promote Safety and Maintain Social Distancing

The advent of the largest and most deadly global pandemic in over 100 years is causing a major disruption in our daily lives.  Some of these changes will be temporary but others will be more permanent in nature.  The 1918 pandemic came in two major waves and there are concerns that COVID-19 will become an annual flu season occurrence. It is reasonable to assume that the need for social distancing and avoiding touching anything touched by others will remain with us even after the virus subsides.   

Computer vision based on A.I. and machine learning is an obvious choice to help. Without computer vision, the alternative is to view hours and hours of imagery and then a very labor-intensive process analyzing the imagery. Computer vision automates image capture and temporal analysis processes. The outputs can include visual dashboards with drill-down capabilities, real-time text, and email alerts. It will emerge as a great training and educational tool as well 


Computer vision is a great tool to monitor social distancing compliance in industrial environments while protecting personal privacy.  The examples below show varying degrees of social distancing in the workplace, from safe and compliant to severe violations of the policy. In the two cases on the right, it is clear that distances are too short to prevent community/people transmission. As a default, Atollogy obfuscates and obscures individual identities, but this can be configured depending on privacy policies.


Processes as simple and seemingly harmless as handing paperwork from one person to another or keying in data on tablets and terminals will now be seen as potential vectors to spread the highly contagious and deadly virus. The less drivers and equipment operators are required to key in data or fill out forms, the greater their personal safety. There are indicators that COVID-19 and related viruses can live up to 72 hours on surfaces, so the fear of physical contact is bound to continue even after infection rates subside.1

Computer vision can not only avoid virus transmission by paper but improve the efficiency of work centers.  In one customer example, operators enter work order data on personal tablets and computer vision is used to validate the start and end times for each job and thus calculate efficiency. It can also be used to count quantities produced in some cases. 

Some of Atollogy’s yard and transportation clients handle several thousand bills of lading (BOL) annually to document the movement of freight shipments. A bill of lading (BOL) works as a receipt of freight services, a contract between a freight carrier and shipper and a document of title. In a typical process, each BOL could be handled by four to five people over one or two days, each of whom could act as a virus vector. Computer vision can replace paper BOLs by capturing vehicle identification, tare and laden scale weights, with entry and exit time stamps.  Where a paper is still required, the BOL can be automatically scanned in an imaging station that uses UV light to enhance safety.

In this example, each document is automatically scanned when it is placed in the box. Unlike a simple scanner, optical character recognition captures quantities, dates, and other critical information. A UV light can be added as an extra layer of safety.  

These boxes can be adapted for exterior uses as well. One such use would be to scan BOLs so drivers do not have to leave their trucks and employees need not handle the paperwork.


Several of Atollogy’s clients require personal protective equipment (PPE), a term that has now entered the global lexicon due to COVID-19. Computer vision can easily and quickly alert management to safety violations. With fewer and fewer middle managers to oversee operations, growing labor shortages, and high employee turnover rates, automating the detection of PPE violations will become more critical.


Re-opening our factories as soon AND as safely as possible is critical to our economic recovery.  We are already seeing the closure of some essential production in our food value chain due to workers in these facilities contracting Coronavirus.  Introducing technology to the workplace that enables industrial workers (who can’t do their work over Zoom) to get back to work safely is a pressing need. Computer vision based solutions deliver remote management capabilities to reduce the number of workers on the floor, and social distance monitoring to keep these front line workers safe – both critical to a sustained response to the pandemic.

The National Institute of Standards and Technology (NIST), part of the Department of Commerce, recently predicted computer vision will be the next revolution in technology: “In a world full of cameras, video understanding – the ability to accurately interpret what is happening in the footage – is likely to become the next revolution in data analytics. Researchers expect to see a whopping 45 billion cameras globally by 2022, and economic analysts expect the video analytics market value to exceed $40 billion by 2023.”2  This equates to five (5) or more cameras for every person on earth. 

NIST estimates that 30 million security cameras are collecting data in the US arguing that the missing link is an affordable video analytics system that can automatically and flawlessly understand events as they unfold in a live feed.” Although there has been a growing investment in computer vision analytics of video, the field remains relatively underdeveloped in the context of public and industrial safety. With over 100,000 deaths globally, COVID-19 is likely to permanently change our social fabric and work environments. Affordable and easily deployed computer vision systems will replace the billions of passive camera systems now operating worldwide. Cameras using computer vision can be a vehicle for much improved public and private safety. As their use grows, debates will continue over violations of personal privacy, but in the great majority of applications, the good should always outweigh any privacy concerns. 


  2. file:///C:/Users/t/Desktop/Enhancing%20Public%20Safety%20Video%20Analytics%20with%20Computer%20Vision%20and%20Artificial%20Intelligence%20_%20NIST.pdf   Stacey Trunnell, NIST, November 2019


Anthony Tarantino, PhD

Adjunct Professor, Santa Clara University – Lean Six Sigma and Supply Chain

Six Sigma Master Black Belt, Certified Scrum Master, CPIM (APICS), CPM (ISM)

Senior Advisor to Atollogy