The idea of analytics in security applications is not a new one. In fact, some in the industry have been (over)promising its capabilities for years.
“AI and deep learning were launched around five years ago with much hype rather than an educational approach, which left both systems integrators and end users a bit confused on the differences between AI, machine learning and deep learning,” says Jason Burrows, Western sales director at IDIS America, Coppell, Texas. “Some early offerings disappointed as they were launched prematurely, before engines were fully trained and able to recognize objects reliably and accurately. The cost and complexity of early-to-market AI applications made users question the value of deployment, configuration and operator use. This was compounded by privacy concerns and a greater storage burden. As a result, many end users were reticent, and systems integrators were unclear on which solutions would be best suited to meet their customer’s needs.”
Many of these concerns are quickly fading, though. IDIS’ 2020 global survey of more than 700 security integrators, consultants and end users showed that more end users had already begun adopting deep learning or had short- to medium-term plans to do so. The survey also showed that both integrators and end users had a clearer understanding of the main security challenges deep learning-powered analytics can overcome.
SDM’s 2021 Industry Forecast Study showed similar changing attitudes. Of those surveyed, 45 percent said they currently offer video analytics, while 26 percent are planning on offering them in the next one to two years, 9 percent are planning on offering them in three to five years, and only 20 percent have no plans to offer them. In addition, 55 percent of security professionals surveyed expect revenue from analytics to increase over the next year.
“What we saw last year was of course an exception, but it also sparked new use cases for analytics — and there were a bunch of them,” says Florian Matusek, product group director, video analytics, Genetec, Montreal.
Many of these use cases went beyond what we typically think of as security, expanding the usefulness of traditional security solutions.
“Cameras themselves and video analytics have really crossed into being a powerful technology that goes beyond just security,” says Will Kelso, vice president of enterprise business intelligence at Interface Security Systems, Earth City, Mo. “Now you’re able to get real positive change throughout an entire organization, so not only are you able to share the spend, but make a big impact across the whole institution.”
Rob Merchant, president of MTS Intelligent Security Solutions, East Brunswick, N.J., featured on this month’s cover, has been an integrator of security analytics for more than 15 years, deploying license plate recognition (LPR) and face detection systems as early as 2006. “Algorithm maturity, application of AI and deep learning and more competition in this space [have changed in the past year],” Merchant says. “I think advances in computing software have allowed these complex algorithms to be able to be scaled to run on more commodity-based platforms. You are seeing these analytics work in more relevant conditions. Aside from LPR and other analytics that have been around for a while, most of these had very niche roles. They are becoming more affordable and more applicable across a greater user base.”
Over the past year, awareness and understanding of AI has increased, both in regards to machine learning and deep learning, says Ryan Gregory, director of solutions and services at Axis Communications, Chelmsford, Mass. “Accordingly, many end users have heightened expectations around the performance of analytics, mostly based on trends and experiences in the consumer electronics space. In many instances, people have come to rely upon these analytics in their daily lives — speech recognition for audio commands or text conversion; optical character recognition for extracting data from print; and facial ID to unlock computer devices. These more common and practical uses have created greater interest, expectations and potential use cases for analytics.”
Aaron Saks, product and technical manager, Hanwha Techwin America, Teaneck, N.J., says there are two pieces to the puzzle that explain why analytics have grown so popular: COVID-19 and AI. “With COVID-19, all of a sudden there was a shift to different people doing different jobs — maybe you had someone watching a camera at a desk, but working from home they’re doing it differently. Also, COVID-19 called for specific types of analytics people wanted to see — looking for face masks, social distancing, whether there’s a car in the parking lot.
“The bigger picture regardless of COVID-19 is the whole AI part. It’s only been these past couple of years that the cost, performance and processing power for AI has been effective and efficient. AI has been around for years, but now we’re in a place where AI can be cost-effectively used.”
A large part of this increased performance and decreased cost is due to the chip size, says J. Matthew Ladd, president of The Protection Bureau, Exton, Pa.
“Technology is able to advance because of the ability to put more analytics onto smaller and smaller chips and concurrently the power and the capabilities of the chips increase,” Ladd explains. “When analytics first came out, you had to connect the camera to a computer — now the camera is making the decisions itself.”
Aaron Simpson, president and chief technology officer at Stone Security, Salt Lake City, Utah, SDM’s 2020 Systems Integrator of the Year, says that analytics are finally getting to a place where he feels comfortable vouching for them. “We have been very cautious with analytics over the years. We have watched many companies over-promise and under-deliver, and often go out of business. That said, we are very encouraged by the advancements in analytics and have adopted many solutions in the past year or two that are delivering impressive results.”
The Many Use Cases of Analytics
Blaine Frederick, vice president of product at Alcatraz AI, Redwood City, Calif., likes to think of security analytics as belonging in two distinct categories: real-time analytics and trend-based or forensic analytics.
Real-time analytics encompass a variety of applications such as face authentication, instances when an item is left behind or a speeding vehicle, Frederick explains. “These types of analytics provide information to systems and users who are making decisions on the fly. They are a key component in any automated system or in security operations centers where there is simply too much data for an individual to process manually.”
Trend-based analytics provide system administrators with insights into what is happening in facilities over the course of time, Frederick continues. “This can be anything from heat mapping in a retail store to vehicle counting at a highway toll booth. Analyzing room occupancy has gained significant attention over the past year as companies plan their return to work after the global pandemic.”
Steve Washburn, executive director of business development for Convergint Technologies, Schaumburg, Ill., adds, “As businesses continue to prioritize digital transformation to attain real-time data and behavioral analytics, more use cases continue to come to fruition. For example, computer vision has unlocked data and analytics that businesses have leveraged to enforce workplace safety initiatives, such as PPE detection, hardhat usage, protective eyewear, gloves or safety vests. Additionally, computer vision technology is flexible and can be deployed across nearly every vertical market, allowing additional use cases to evolve over time.”
At their simplest, analytics make security personnel more effective, says Paul Garms, director of regional marketing, video systems, North America, Bosch Security Systems, Fairport, N.Y.
“When thinking purely about using video surveillance for the protection of people and assets, analytics make security personnel more effective by raising their awareness and ability to react to system events,” Garms explains. “In a manned system environment, this adds an extra layer of protection by providing alerts to potential security risks before or as they occur — such as detecting loitering in a parking lot or a perimeter breach after hours.”
Conditions that video analytics can be programmed to alert on include line crossing, illegal parking, loitering, speeding and color matching.
False alarm filtering is another hot analytic offering in the security industry right now. Calipsa, Ashburn, Va., provides one such solution.
“Our software platform identifies the live movement of humans and vehicles within the view of the surveillance cameras,” says Calipsa Chief Revenue Officer Brian Baker. “Nuisance factors, such as windblown foliage, are filtered out, significantly reducing false alarms and improving a video monitoring agent’s response times to genuine security threats while improving their overall efficiency.”
One of the factors that makes analytics so appealing is that it allows for a proactive response, rather than a reactive one. “When it comes to security applications, analytics can offer many advantages by serving as a force multiplier, enabling proactive response and providing operational efficiencies,” Axis Communications’ Gregory says. “For example, in the past, security personnel were required to monitor multiple live feeds and react to incidents or scour hours of footage from a forensic standpoint. Today, video analytics are helping end users quickly identify objects or sounds in real-time and respond in a proactive manner by dispatching live personnel or even triggering a sound recording to deter trespassers or undesirable behavior.”
“Customers want to know where people are within a building and to predict situations before they happen,” says Sheeladitya Karmakar, global offering leader for Honeywell Commercial Security, Atlanta “We can create reports on how many zones an individual has accessed in the last 24 hours and are developing the ability to predictively tell security teams where people will be in the next 20 minutes. This will help security teams get ahead of possible issues. For example, we can leverage data to know that many people exit the building around a certain time, so we can preemptively open or unlock doors to allow for a more efficient process.”
While cameras tell you a story, analytics help you write the story.
“Analytics can be used in multiple ways to help safeguard people, places and property,” says Devarshi Shah, CEO and founder of Lumeo, a custom video analytics platform based in Oakland, Calif. “Cameras can only tell a story, but analytics can help security react and respond to any given incident. For threat management and incident response, they can let first responders know where an individual is located and allow for proactive response at venues with respect to crowds or alert when a secured area has been breached. Beyond that, analytics can also be leveraged for compliance and better operational planning, like where and when to deploy security. There are countless ways in which analytics can be leveraged within the security industry.”
John Nemerofsky, chief operating officer at Sage Integration, Kent, Ohio, says the analytics application his client use the most is multi-camera search.
“When analytics first came out, it was about law enforcement being able to do investigations quicker with a single camera; but multi-camera search is the most popular analytic we’re using,” Nemerofsky says. “Another thing we’re using as people move back into corporate headquarters is watch list rules. It could be for bad guys, or it could be used for VIPs so when they arrive at the building, everyone can stand at attention and make a good impression — we’ve used that a lot.”
Ladd has also seen a number of applications using people tracking and the patterns that come from that. “Look at contact tracing — you’re using this capability with the camera. Analytics can be used to track people from one camera to another. In an emergency you can track someone as they walk through a building, such as an active shooter. The camera locks onto a person and tracks them until they reach the next camera, and so on. In the retail market, it’s been talked about for a long time. Cameras can track shoplifting, but also work with marketing departments to give them data on displays and how people move around the store. The camera helps the department decide where to place certain items in the store and helps them analyze how that impacts shoppers’ time and buying decisions.”
Analytics can even detect cyber and terrorist attacks, Baker says. “Law enforcement and homeland security officials use analytics to monitor social media, phone communications, travel records and many other data sources to spot potential crimes or terrorist attacks before they occur.”
Too many options to choose from? You could always mix and match.
“A la carte analytics seem to be gaining popularity,” Stone Security’s Simpson says. “As mentioned before, most analytic platforms have forced customers to deploy a single solution across their entire application and require costly on-premise resources to do so. Many analytics are only needed in a few locations, and most customers would benefit from a variety of analytics.”
A Limitless Future
Right now, it seems we are only at the beginning of what analytics will ultimately be capable of.
“Artificial intelligence and its subsets, machine learning and deep learning, massively expand the capabilities of today’s analytics,” Baker says. “However, training these analytics is a time-consuming and costly process. For example, training a deep learning platform to accurately detect the differences in movement caused by a human or stray dog requires exposing neural-node networks to millions of photos of humans, making corrections each time the software makes a mistake. Over time, the platform will accurately predict the presence of humans or whatever else it has been trained to recognize.”
Supply chain shortages caused by the pandemic are stalling progress a bit for now, Hanwha Techwin’s Saks says.
“[In order for analytics to progress] the cost needs to go down, and right now with post-pandemic chip shortages and shipping delays prices aren’t going down; but generally they do,” he explains. “What we’re going to see is analytics becoming more commonplace. We’re going to get the camera optics better optimized and get the best wide dynamic range. Often AI would be included in the premium lineup of cameras, but as product lines mature we will see AI cameras becoming more of the feature sets you typically expect on all of your cameras without limitations.”
Since many companies require high compute power to run analytics on edge-level devices, according to Karmakar, the decreasing cost of compute power should be helpful. “This is happening gradually, not drastically, but it’s happening. This is why we have access to so many features on our smartphones today that came along in the early 2000’s. The same will happen for these analytical servers, which will make higher compute power more accessible.”
While lower prices are necessary for growth, they’re also inevitable.
“In order for analytics to reach their full potential in security applications, prices will need to drop, while technologies and their capabilities will need to increase,” Ladd says. “I’ve been in this industry for 46 years, and everything continues to grow and get better and faster. When cameras with analytics were first introduced in the security industry, they cost thousands of dollars, but now the price is a fraction of that. I’m always impressed with how things continue to develop and go further — it’s up to the processing capabilities and how you send out the information. Everything is wireless and yet there is so much further to go. Technology will continue to grow because the desire for smarter and faster technology will always be there.”
Lower costs and improved technologies go hand in hand, Merchant explains.
“I think advances in computing hardware have allowed these complex algorithms to be scaled to run on more commodity-based platforms,” Merchant says. “You are seeing these analytics actually work in more relevant conditions; aside from LPR and other analytics that have been around for a while, most of these [other applications] had very niche roles. They are becoming more affordable and more applicable across a greater user base.”
Analytic companies that don’t improve the cost and ease-of-use will likely not last much longer, Simpson says. “Many traditional analytic companies are currently working to reduce significant hardware resource demands and leverage the untapped edge capabilities. I believe we will see some of the more proactive companies adapt to the ever-changing environment while others will remain stagnant and fall by the wayside.”
As these technologies become more accessible, security integrators will need to become educated on how they work — or they could fall to the wayside, too.
“Increasing awareness of the full capabilities of analytics is essential,” Garms says. “Integrators need to get trained on the new possibilities for video analytics that are enabled by AI, and they should look to manufacturer training courses to assist with this. Understanding the various use cases where analytics can be applied is also important. This will help integrators identify situations where IoT video systems will be beneficial for their customers.”
And since analytics are ever-changing, education will have to be continuous — not just a one-time course.
“For analytics to reach their full potential, improved education and training is required,” Gregory says. “A well-informed value chain — from developers and systems integrators to end users and even the public — can properly set expectations and provide greater understanding. Also, while major strides have been made around AI training models, an even greater emphasis on continuous improvement will benefit existing applications and spark new applications.”
Nemerofksy says that new unique use cases will push the technology to advance. “It’s going to take the end user continuing to come up with challenges and problems they want to solve. … It’s going to take that security practitioner to keep challenging integrators.”
In reality a lot of the technology is already here, Interface Security’s Kelso says. “It’s ever-growing, but the selection of a right partner to guide you through the human element and also having that ease of set up and being able to take these data sets and make sure they’re actionable is key. And there is an opportunity with the technology in place to gain a lot of great insight.”
Quality guidance is especially important in an industry like security, points out Jonathan Moore, vice president of product management at AMAG Technology, Hawthorne, Calif. “The security industry is a tough market because a security failure can carry serious implications, opening the organization to unnecessary risk. Analytics solutions are improving; however, the security market is typically looking for perfect or near-perfect solutions. For analytic solutions to reach their full potential, they need to operate at a very low failure rate while keeping the cost of the solution (and the supporting servers and technology) at an acceptable level.”
Kai Moncino, global business development manager, security, Teledyne FLIR, Thousand Oaks, Calif., says, “The goal for any analytics software is a detection rate approaching 100 percent and a false alarm rate approaching 0 percent. While this is nearly impossible in real world scenarios, there are some analytics that are closer to these metrics than others. Looking forward, the strongest competitors in the video analytics space will be those that deliver consistent detection rates and are able to reduce false alarms.”
The Time Is Now
Like many other technologies, the adoption of analytics sped up during the pandemic, as businesses were forced to accelerate their digital transformation initiatives to survive, Convergint’s Washburn says. “Now, an abundance of data can be applied for business intelligence purposes, allowing them to thrive. We’ve reached an inflection point where now, the technology is so far advanced, it’s time to maximize use and unlock untapped potential. The limitless flexibility behind computer vision’s AI offers new opportunities, leveraging ground truth and sample data to train machine learning models to recognize specific elements or behaviors.”
Now, security professionals just have to get in on the action — they already have the assets.
“As an industry, we have a huge asset we are sitting on — there are over 1 billion security cameras deployed worldwide,” Lumeo’s Shah says. “Most of these cameras are just recording devices. But when you start thinking about the wide range of use cases you can solve for using these cameras and AI-driven computer vision, you break out of the silo that these cameras are in today. You start to realize their true potential: to solve business problems across various domains, improve customer experience and open up new revenue streams.”