The Power of Deploying Video Analytics in the Cloud
As technology migrates to the cloud, so do analytics, offering more benefits than you might think.
Technology migration can be tricky to predict, but it seems safe to say most video analytics will be deployed in the cloud within the next couple years or so.
Already, a good number of companies are seeing the benefits of analytics, whether in the cloud or at the edge. Chris Brown, vice president central stations, SureView Systems, Tampa, Fla., says out of the customers SureView works with, which is a majority of the mid-size to large central stations, he sees probably 65 percent of customers engaging in some type of analytics.
“I think that number is growing rapidly as we speak,” Brown says, “and I would say given two years, probably almost every central station that has a camera coming inbound will be relying on some type of analytic, either at the edge or in the cloud, in order to generate that signal and create the event.”
And by the time most come to adoption, he adds, cloud will be how most analytics are deployed.
While it is fairly common among large, enterprise systems, James West, CEO and co-founder, Manything, Oxfordshire, U.K., says the majority of installations he sees are still using local recording. “A hybrid model (for example, doing some on the camera and some in the cloud) is necessary in the short term to get the costs in the right ballpark to increase adoption,” he adds. (See sidebar “Preparing for a Storm,” page 63, for more on hybrid systems).
There are still some hurdles to overcome before the majority of analytics are deployed in the cloud. One of those hurdles is bandwidth expense, says Ken Francis, president, Eagle Eye Networks, Austin, Texas. “An effective cloud system requires bandwidth to get the video off site quickly and efficiently.” But as technology continues to improve and bandwidth expense to customers continues to be reduced, he explains, it’s only a matter of time before all customers can consider the benefits of a cloud system.
While this might not be the greatest news for camera manufacturers selling analytics, it is the reality of the current trajectory and is good news for security integrators and central stations thanks to the benefits deploying analytics in the cloud affords everyone, from integrators to end users.
ABLE TO PROCESS
The first advantage of cloud deployment is the unlimited processing power deploying in the cloud allows. Analytics have advanced exponentially in a very short amount of time, and much like the explosion of computing power of PCs and personal devices in the past decade or two, the advancements in this area will only continue to snowball.
But just providing analytics isn’t enough; the analytics need to be highly accurate so that by the time an alarm gets to the monitoring center, it is actionable, Brown explains. The advancements in analytics require a lot more processing power than in the recent past, and analytics on the edge are limited by the processing power of the chip that the camera has.
“You can create as much processing power as you desire in the cloud to process that image,” Brown says. “You can stream that image in and process it in a place where you can put real power against it.”
This is a train companies will want to get on sooner rather than later, West says. “The technology is going to get very good very quickly, and you risk getting left behind if you don’t start building up some experience by riding the slipstream of large organizations with enormous datasets on which to build their data models.” He offers Amazon and Google as examples of companies running developer programs for their image and video analytics.
Another aspect of that processing power is the ability to make sense of a large amount of data. West says most video being captured by the security industry today is being added to what he describes as a “chaotic library where books are stacked on shelves in date order only, unlabeled, making them hard to find.” He says the main opportunity for analytics in the short term is to improve video labeling to make retrieval and alerting more reliable. This could make mountains of video footage that is not associated with security threats suddenly become valuable to the consumer.
UPDATES FROM ON HIGH
Another advantage of deploying analytics in the cloud is the ability to quickly and globally update an entire analytics suite in one transaction.
Unlike alarm panels or relays, analytics are ever developing, and every company that produces them has its own algorithm against their own imaging to better improve and let their analytics learn. Without the ability to push mass updates from the cloud, Brown explains, you’re trying to push updates to the algorithms that produced the alert into a very light, small edge device. “It’s time consuming; it’s inconvenient for the customer; it takes the customer down while the service call is being done — it’s a very clunky, old way of doing things.”
A true cloud system eliminates the need of specialized, on-site analytics systems that need to be integrated and maintained, Francis says. “If planned properly, a customer can practically eliminate the need for truck rolls to turn on/turn off or modify analytics settings. Most importantly, a cloud system customer can start and stop billing at will, enabling short bursts of at-will usage.”
In addition to making global updates, companies can make regionally focused updates such as filters. “Certain environments sometimes require certain types of filters,” Brown says. “For instance, in Texas it rains sideways. So being able to apply a patch or filter to geographically positioned cameras, you can make adjustments like that that are [more] specific to environmental threats or changes, as well.”
He says this gives analytics providers maximum flexibility to be able to update and create new offerings inside an integrator’s package as analytics develop and technology continues to develop.
Doing things this way also allows for various RMR opportunities, giving a company the flexibility to offer, in addition to security analytics, non-
security analytics, such as people counting, as add-on features that can be easily plugged in, pushed to the cloud and offered as a service. This also adds to the flexibility to become product-agnostic.
A WORLD OF POSSIBILITIES
Deploying analytics in the cloud allows a provider to truly sell analytic power, and not just a device, Brown says. “You start to allow people to use analytics they historically couldn’t and you gain customers that wouldn’t roll a truck or pay a capital expense to change out a fleet of cameras,” he says. “We have several partners that have the ability to run analytics inside the center in the server, on the cloud. And because they can do that, they are not restricted by brand or type of camera. And if the image meets the right quality, no matter who the manufacturer is, all of a sudden you can start to add analytics, whether it be security analytics or business data analytics, because all you need is the image.”
With companies already spending huge amounts of money to deploy cameras, this allows the flexibility to add analytics to any camera, especially if they have a system of cameras without analytics built in.
And while many might fear cyber threats associated with the cloud, it is important to note that world-class cloud service providers offer the highest levels of physical security for their datacenters since they have to comply with regulations such as SOC 2, ISO 27001, HiPAA and PCI, says Oktay Yildiz, product line manager, Stratocast – Genetec Inc., Montreal, Canada. “Often, their security and redundancy practices are extremely difficult to replicate by a SMB due to the costs involved. Furthermore, mundane IT tasks such as infrastructure maintenance and patching are done in a timely matter, ensuring the security of the services provided.”
In fact, Yildiz explains, it is a step in implementing an appropriate backup plan by having data stored offsite, in multiple copies. “It could also help in your recovery plans in case of a disaster,” he adds.
Another fear some have is that traditionally cloud-based analytics have had a significant lag compared with edge-deployed analytics. “Historically some cloud-based analytics services would take a little bit longer for those alarms to get into the monitoring center because they have to go from the edge to the cloud,” says Jason Caldwell, business development manager, SureView Systems, “and then be pushed into whatever software automation a central station might be running. So sometimes you would be getting a video clip of an event, and it might be 20 or 30 seconds old, which in a security threat situation, that could be a long time.”
But this is quickly improving, Caldwell says, and Brown agrees. As customer demand drives more analytics adoption, the power and accuracy of cutting-edge analytics, along with all the other benefits, makes this a sensible option.
Deploying analytics in the cloud gives companies the ability to add analytics to any camera system or just to make the analytics better — more flexible, more affordable and more reliable.
Preparing for a Storm
One hesitation some might have about deploying analytics in the cloud is the fear of putting all their eggs in one basket, as it were: If the cloud goes down and that’s where all your analytics are deployed, then you have no analytics for a time.
That’s where a hybrid model can be attractive, says Chris Brown, SureView Systems. “Some companies are doing their analytics in the cloud, but they’re also running a light analytic in the field. There are manufacturers out today that are running a light analytic at the edge that is allowing the brand name device to see the opportunity for a threat. It’s not always accurate, but it’s efficient,” he says.
Brown explains if the analytics program on the edge device detects something, it generates a video clip and then hands it to the cloud. The cloud does another process against it, which is much finer and sharper to determine if that alert is going to move to an actionable person — whether it is the homeowner, the business owner or a central station — to see whether that clip must be looked at. If it is verified as an actual alert, it moves. It all happens in a millisecond, Brown explains.
Brown says in a hybrid model, having some type of light analytics running at the edge allows users to still have security on the perimeter in the event that the cloud array goes down. “It gives you a nice safety net of at least getting protection. Although there are going to be more false alarms to deal with, you’re still protecting customers.”
Agent Vi is a video analytics provider that offers such a hybrid. “Agent Vi’s cloud solution uses a distributed processing architecture whereby some of the analysis is performed on the edge, and only low-bandwidth data is uploaded to the cloud,” says Zvika Ashani, the company’s CTO and co-founder, “thereby enabling a practical implementation of highly accurate video analytics deployed in the cloud.”
Ashani says this implementation is already commercially deployed around the world.
Leave the Decisions To the Machines
With machine learning and predictive analytics, the detection of human behaviors and intentions could be the next breakthrough, says Oktay Yildiz, Genetec Inc.
“Most monitoring stations heavily rely on a person to ascertain if analytic events could imply the negative intention of a suspected individual. With machine learning, a scoring model could be applied on that individual. Incremental improvement in developing more efficient algorithms could continue to reduce the cost of processing analytics,” he says.
Also, with deep learning and some artificial intelligence (AI), the amount of configuration parameters on analytics will be greatly reduced. Ironically, Yildiz says he expects to see more analytics done directly on an edge device or on “smart” cameras, rather than solely in the cloud with the reduction of configuration parameters.
For more on cloud video surveillance, visit SDM’s website where you will find the following articles:
5 Ways to Increase RMR With Cloud Video Services
An Integrator’s Guide to Pricing and Selling Cloud Video Storage
How the Cloud Can Add Value to RMR Using Cloud-Based Video Storage
The Continuing Evolution of the Cloud
Where’s the Gain in Security Cloud Computing?