The vaccine rollout is being met with lifted COVID-19 restrictions inside buildings and restaurants, but this change presents a new challenge to business owners -- managing increased occupancy, while still abiding by safety restrictions. Businesses that find themselves exceeding occupancy could face fines, citations, and license suspensions.
One increasingly prominent solution employs 3D counting and tracking cameras that monitor occupancy, foot traffic, and flow inside brick-and-motor locations. Regular 2D cameras and traditional counting techniques are not accurate enough. However, depth-sensing 3D cameras can provide real-time updates that increase counting accuracy by an estimated 5% to 8%, according to a spokesperson for one 3D company I spoke with, Orbbec. That difference in accuracy is crucial when limiting traffic is a matter of health and safety.
Of course, privacy considerations and technology adoption issues will tell the tale when it comes to 3D camera technology. I know Orbbec for their technology's use in robotics, but the people counting application left me intrigued. To find out more I reached out to David Chen, Co-Founder and Director of Engineering at Orbbec.
GN: What are the key differences between standard visual spectrum 2D cameras and 3D camera technology?
David Chen: The major difference between standard visual spectrum 2D cameras and 3D cameras is that 3D cameras provide direct and true distance information while a 2D camera cannot. In addition, standard visual spectrum 2D cameras see through visible light within a very large wavelength coverage, meaning the quality of the image is highly dependent on environmental lighting and illumination.
Orbbec's technology in particular uses a NIR (near-infrared light) source and a narrow-band optical filter. The narrow-band filter helps to minimize the influence of environmental illumination (commonly known as noise); on the other aspect, the active light source, passing through DoE (diffractive optical element) can form different patterns that can be used in the 3D reconstruction.
These features allow the 3D cameras to accurately perceive depth and can be used as a navigation solution for self-propelled robots in healthcare, food service, warehouses, schools, offices, and other places populated with people or other objects that are on the move. 3D cameras and sensors are also better suited than 2D for facial recognition at high-security locations like payment kiosks and ATMs.
GN: Why are depth-sensing cameras better for counting people in real-time than 2D cameras? How do accuracy rates compare?
David Chen: 2D cameras can be easily fooled, leading to inaccurate results. This typically happens in the form of tailgating or piggybacking, when two or more people are grouped and trick the camera into thinking they are one.
3D imaging systems typically include multiple RGB or IR (infrared) cameras that add distance and/or depth information to each pixel in the image. This allows the distance between objects to be accurately measured in a scene, all in real-time.
With the depth information taken from a 3D camera, programmers can then accurately track people with a simple algorithm, especially useful in complex scenarios or locations, like a crowded entrance. As brick and motor locations need to keep a precise count of the number of people entering their locations, the difference in accuracy can be critical. 3D cameras for people counting increase accuracy by 5% to 8% compared to 2D technology. According to our partner and clients, some 3D-based counting systems can reach up to 99% accuracy.
In restaurants, 3D cameras can be programmed to count, track, and log not only the number of people in a room but their position, movement and grouping as well. 3D cameras can also be programmed to scan and recognize objects, like the number of similar items at a grocery checkout or on a cafeteria tray. This can be an invaluable tool in shipping, packaging, manufacturing, and warehousing operations.
GN: Privacy is obviously a concern in any commercial application. How can deployments meet the need of the pandemic while maintaining privacy?
David Chen: Compared with 2D methods, 3D increases privacy because it only shows depth information if needed. Around the world, 3D facial recognition is in daily use at millions of locations. Customers actually prefer 3D facial recognition in commercial and safety applications for its speed and convenience.
For people counting and tracking solutions, 3D technologies (and their accompanying algorithms) don't rely on traditional 2D images or videos. 3D cameras only record sparse 3D point clouds (points within a 3D image) that can be recognized only by computers. Making 3D cameras an even more effective way to protect privacy
All 3D cameras from Orbbec have traditional RGB cameras within but we provide customers the option to use it or not. And we are also able to help remove the RGB camera at the customer's special request for enhanced privacy requirements.
GN: Where is Orbbec's technology currently being implemented related to the pandemic? Can you describe the kind of setup involved for end-users?
David Chen: 3D cameras are helping to ensure that safety protocols are being followed in quick-serve establishments, stores, bars, and restaurants. When stationed high in a ceiling, for example, 3D cameras can track occupancy, flow, and clustering. It can automatically warn managers when social distancing measures are not being followed.
3D cameras make contactless ordering possible via eye-gaze tracking or "air pointing." Stores are upgrading ordering kiosks with 3D cameras to order and pay without touching. Elsewhere, sanitizing robots use SLAM (Simultaneous Localization and Mapping) technology to avoid obstacles; robots can sanitize entire hospitals without exposing cleaning staffs to infection.
The top five applications for our 3D cameras include body/people counting; biometric payment -- commonly in the form of facial payment technologies; air pointing; a tool to help maintain social distancing (i.e. keeping safe working distances on job sites or ensuring proper distancing for people in queues); embedded into robotics technology -- such as delivery robots or sterilization robots found in hospital.
GN: How competitive is the depth-sensing space for COVID-19 applications? Is this a sector that's seeing a lot of competition currently?
David Chen: The depth-sensing technology is replacing and combining with the 2D technologies gradually. We do see many companies are seeking to integrate people counting and depth-sensing technology every day. Not only does this technology help manage social distancing, but it helps increase sales and shopper convenience in the retail world.
A May 2020 CapGemini Research Institute study on contactless customer experience found that 84% of U.S. consumers expect to increase their use of touchless technologies during the COVID-19 crisis, to avoid interactions that require physical contact; of those, 55% expect to use touchless technologies even after the crisis ends. Fifty-two percent of respondents said they prefer facial recognition for authentication at retail stores, banks, airports, and offices during the COVID-19 crisis.
Orbbec's partner Moptar has found people tracking can visualize the movement of all customers in the store and verify the effectiveness of various measures. For example, 3D camera technology allowed an electronics store to verify the effectiveness of in-store advertising and increase the number of visits to the target sales floor by 1.4 times.
So, when companies and consumers really begin to understand the advantages of 3D cameras, we expect competition to only grow.