A confluence of factors is causing artificial intelligence (A.I.) to make serious inroads into integrated physical security technologies. These include refined computer vision technology and faster processing speeds, improved analytics and data procurement, cloud computing, enhanced machine and deep learning algorithms, and edge computing. It’s the perfect storm—in a good way.
A.I. makes systems integrations smarter. As an example, when access control data yields how many people are in a building, you can use that information to optimize energy management and lighting systems or to facilitate emergency evacuation. When the last person badges out, the system can revert to nighttime mode or shut down for greater efficiency.
In every physical security product category, including video, intrusion, access control and others, A.I. is expanding potential applications. A.I. data effectively saves time, minimizes errors and takes systems beyond security into safety, facilities management, occupancy controls and other business operations.
A.I. brings eyes to video
In surveillance, A.I. solves the problem of managing hundreds of video feeds and conquers false alarms common in capturing outdoor activity. Monitoring centers no longer watch video 24/7 and instead rely on alerts created by A.I. cameras that leverage machine learning and edge computing to detect and react to anomalies, continually adapting to refine detection parameters.
“With staff shortages and operations and security professionals stretched to do more, being able to use analytics truly gives them an extra pair of eyes that never sleeps,” said Aaron Saks, Hanwha Techwin’s senior product and technical training manager, Teaneck, N.J.
“Users can get alerts in real time, as well as conduct smart searches post-event. This enables teams to be proactive rather than reactive,” he said. “Some video management systems send alerts to user’s mobile phones so they can react from wherever they are.”
Saks said another benefit is that A.I. cameras receive updates that increases their utility with plug-ins and third-party developers.
“This type of open ecosystem means a camera investment can mature over time, like any computer that can be taught to do new things. Business and operations intelligence brings what was just a security camera into the realm of a smart IoT sensor and a completely different type of system,” he said.
A popular and refined application for A.I. is intelligent automatic license plate recognition and the ability to recognize make, model and color of vehicles while showing traffic data. Cameras support real-time viewing, event monitoring, forensic search and a range of widgets and graphs that display statistics.
“With multiple types of classifications, including car, SUV, van, light commercial vehicle, truck, bus and motorcycle, users can search based on a number of attributes including date/time, license plate, photo, type, brand, model and color,” Saks said. “In addition to displaying plate and vehicle specifics, A.I. edge-based applications also support black/white list event notification. As a standalone system, the application can support up to 10,000 detailed records, including the images of plates and vehicles, on the LPR camera’s internal SD card.”
Saks predicts further customization of A.I., tailored to the user and the type of detail they want to capture. For example, users may want to be notified of a specific type of truck entering the facility, PPE detection or even man-down events.
Edging into access control
One of the most significant developments in access control in recent years has been the use of edge computing, said Blaine Frederick, vice president of product at Alcatraz A.I., Redwood City, Calif. Alcatraz A.I. replaces access-control badging by leveraging facial recognition, 3D sensing and A.I.
“A.I. allows users to do more with less. It removes the tedious, time-consuming and human-error-prone tasks to be delegated to computers, freeing up resources and allowing for more data-driven decisions while increasing productivity and efficiency. A.I.-based analytics is becoming widely utilized to detect patterns and alert resources of any aberrations or inconsistencies,” he said.
Another application of computer vision and machine learning in access control is the devices’ capability to make decisions around physical security based on traits beyond facial identity.
“As a result of the pandemic, for example, many customers want to prevent building access to unmasked individuals, even if those individuals would otherwise be granted access based on facial authentication,” Frederick said. “The same situational analytics can be applied to other scenarios (such as the use of hardhats and safety glasses) in which customers want to consider both identity and compliance when establishing their standards for access control. In these ways, the technology has gotten smarter with new deep learning algorithms.”
When A.I. is deployed for perimeter security, the system makes real-time decisions at the door and provides a myriad of data and analytics. This removes the need for security administrators to comb through massive databases or manually input criteria for access.
“For example, if a suspicious bag is left abandoned at an airport, an A.I.-based video system would detect the bag, notice that this was inconsistent with accepted security protocol and immediately alert administrators of the potential threat,” Frederick said.
“Even apart from security, A.I. allows video systems to trace the flow of customer traffic through a retail store via heat mapping. Store merchandisers and marketing resources can then leverage that data to ensure that products are placed in those high-traffic areas, allowing them to advance key marketing objectives in a targeted and data-driven way. In the future, I believe we are going to continue to see an evolution in the usage of security-related data across a variety of organizations and industries, widening its applicability from physical security alone and crossing over into marketing and human resources,” he said.
Crossing over to commercial
Jean-Simon Venne, Chief Technology Officer and co-founder at BrainBox A.I., Montreal, said predictive, self-adaptive, cloud-based A.I. optimizes energy savings, which increases operational efficiencies and extends equipment life while maximizing occupancy comfort. Venne said BrainBox A.I. uses A.I. technology to proactively optimize one of the world’s largest energy consumers and greenhouse gas emitters: buildings.
“What we have seen in the last five to six years has actually been a wakeup call. A lot of commercial buildings have made investments into HVAC controls deployed across their portfolios. These systems provide better management of equipment, and people are starting to realize they generate a great amount of data—on temperature, humidity and more. That data is normally discarded once the real-time operation is done,” Venne said. “So, the thought is: Maybe we can regenerate that data for other uses. Facebook and Google basically showed us what’s possible with data and A.I. If you have data, you can use mathematical formulas and A.I. machine learning to conduct predictions, and when you open that field there are all kinds of things you can do. In HVAC, that data is already there. Neural networks, at the heart of deep-learning algorithms, just started in the last seven years. Computing and storage costs have gone down. We now have the capabilities to create value with the data from HVAC systems.”