Crowd detection analytics comes in many forms, from edge-based to server-based. Choosing the best solution depends on the environment and need.
Choosing whether to deploy an edge-based or server-based crowd detection system depends on the application and environment. Operators should take into consideration crowd sizes, egress points, etc., and keep in mind that there is no “one size fits all” approach.
Benefits of video, edge-based crowd analytics
Stuart Mills, COO,
Chief Customer Officer, CrowdVision
In terms of cost, video is one of the most effective ways to deploy crowd analytics since it can leverage existing surveillance infrastructure. Using cameras also allows the venue to visually identify individuals in a crowd, if needed.
Tom Hofer, Product Manager at
Senstar, noted that other solutions could offer higher accuracies than video — for example depth sensors and LiDAR process 3D datasets — but these approaches require substantial computing power and would have considerably higher equipment and maintenance costs.
One approach that balances both cost and processing power, is to deploy
GPU-enabled edge devices, Hofer said. “GPU edge devices free up network and server resources by performing the computing outside of the server room but remain camera-agnostic, a key factor in reducing overall costs,” he added.
As one of the evolutions in this space has been the development of multi-sensor solutions, CrowdVision offers both video and
LiDAR solutions through a unified customer dashboard, said Stuart Mills, COO and Chief Customer Officer at
CrowdVision.
Support for both edge- and server-based
Crowd analytics providers implement solutions in different ways, including traditional on-premise implementations and now edge and cloud computing. In fact, to be able to run crowd analytics at scale across many sites, both IoT devices and cloud computing should be used, according to Mills.
The CrowdVision detection engine can be deployed on either an edge device, which sits alongside the camera, or on a designated server within the customer’s IT footprint. The
benefit of the edge device, though, is that it immediately analyses the video stream and converts it into data, which is then quickly transferred to the cloud, Mills explained.
“CrowdVision supports both edge-cloud deployments and on-premise deployments, depending on the preference of the customer. Both methods are equally reliable, and the selection of the suitable architecture is typically guided by the customer depending on their preference and IT infrastructure,” Mills said.
Bjørn Skou Eilertsen, CTO of
Milestone Systems, explained that the benefits of implementing the “right” platform are that it can adapt to many situations, save time and money and makes it easier to solve new upcoming challenges.
As with all software, hardware is required to enable the capability to run the actual analytics. “This hardware could vary from units at the edge of a system like today's powerful IoT devices to on-premise systems or
cloud solutions. Open platforms adapt and scale to most of those situations because of the strong partnerships with hardware manufacturers like Intel, Dell and Aopen and
cloud providers like AWS,” he said.
Taking all this into consideration, Hofer says if you have a method of uniting metadata from multiple sensors, both approaches will work.