Developments in AI await video surveillance consumers

a&s did a Buyers' Preference Survey on around 200 professional systems integrators, distributors/dealers/importers, consultants, as well as installers worldwide. The 200 respondents shared the key product features that are in demand in the current market.

In the video surveillance segment, 55.4 percent of the respondents said that they were likely to purchase and install artificial intelligence (AI) based solutions. Popular applications of AI in video analytics include analytic solutions like facial recognition, people counting, behavior analytics, and license plate recognition.

What customers can expect

Speaking to a&s on this trend, Pieter van de Looveren, Director Marketing Communication Security at Bosch Safety and Security Systems said that today, the continuous advancement of security solutions is being offered with a substantial increase in the level of AI and edge-intelligence.

“A good example of edge-intelligence is built-in video analytics,” van de Looveren said. “The camera should not only be able to distinguish individuals from each other, but also, for example, from vehicles. A good example of AI is facial recognition at airports or other applications. Without the use of AI, it is difficult to easily recognize persons in so-called ‘non-cooperative face recognition’ situations.”

Indeed, facial recognition software solutions based on AI can be used to find and identify missing or wanted persons or objects in more crowded scenes, identify VIPs, or enhance customer experience with special offers, services, and seamless access.

As surveillance video migrates from on-premises/to the cloud, the opportunities for adding AI routines to video becomes far easier. According to Tom Buckley, Co-founder of the startup Qumulex, searching for specific things within the video will be one of the major functions that will become easier through AI-enabled natural-language search queries. Then there are also benefits that go beyond security.  

“Surveillance video can also provide marketing feedback, as AI analyzes user behavior in the video, such as length of time spent at certain retail displays or museum exhibits,” Buckley said. “We’ll see easy app store-like routines to add to existing cloud-based video systems.”

Jeremy Kimber, Global Video Product Management Director at Honeywell Commercial Security added that combining the use of AI with existing analytics may also provide benefits in certain circumstances.

“At the headend, AI enables the synthesis of camera, video and sensor events/alarms for smart decision making and fast response to security issues,” Kimber pointed out. “For operations, AI can effectively reduce false alarms and discard irrelevant recordings so that operators and investigators can save their efforts to focus on events critical to security and the safety of the assets and people they are protecting.”

Kimber added that some AI-based features that will be available in video cameras and servers include facial recognition, object detection and classification, behavior detection and classification, anomaly detection, multiple object and people detection and classification of a scene, quick search of objects, people, vehicles, and vehicle classification (type, make, and model).

 AI to be a norm across verticals  

Customers are now demanding a range of AI functionality at the edge, and in the next couple of years, we can expect this capability to become

the norm for commercial surveillance cameras. According to Joon Jun, President of Global Business Division at IDIS, most reputable manufacturers need to offer trip zones, auto-tracking, and auto-detection on their most popular cameras, especially those for perimeter and other outdoor applications.

“For retail customers, heatmapping and queue detection/management are becoming essentials, extending the value for surveillance beyond just security to marketing, sales, and wider business operations,” Jun said. “For example, store managers can receive alerts to open new checkouts, reduce waiting times, improve transaction efficiency and enhance the whole shopping experience. Heat mapping allows retailers to better position promotions and merchandise based on the amount of time shoppers spend in specific areas of a store. It will identify hot spots and dead zones so store designers can optimize their layouts.”

 The popularity of multisensory cameras

The survey results had also shown that 41.2 percent of customers would opt for multisensory cameras. Commenting on this, Kimber noted that multi-sensor cameras are increasing in popularity because customers can deploy only one device instead of multiple IP cameras when covering extensive areas like warehouses and parking lots.

“This saves on both equipment and installation costs: using one multi-sensor camera instead of multiple single-sensor cameras saves camera

costs, cabling material and manpower costs (pulling one cable instead of multiple cables) as well as configuration time (one IP address instead of multiple IP addresses),” Kimber said. “New developments will bring higher megapixel cameras (over 20 MP), high compression codec support, video analytics, increased combinations of normal and thermal multi-sensor cameras, long-range multi-sensor cameras, higher frame rates at 60/120 fps, and higher Wide Dynamic Range (WDR) at over 150 dB.”

The rising importance of IoT

A lot has already been said about the Internet of Things (IoT), so much so that any more predictions of IoT playing a key role in security could soon become cliché. However, the truth is that given the enormous potential IoT can bring, then the importance of IoT cannot be stressed enough.

“IoT is changing the way we view video security, quite literally,” Jun said. “Now, video security cameras need to be so much more than devices with which to simply capture images; they need to be far smarter than that. Today cameras are becoming an integral part of the vast digital connectivity infrastructure, they need to be shaped into intelligent sensors that have the ability to extract invaluable data to help businesses make improvements in the area of video security. This also applies to the multisensory segment. Cameras are increasingly seen as IoT sensors. Therefore, multisensory cameras with built-in intelligence make sense.”

IDIS is seeing a demand for multi-sensor cameras for quite specific applications, such as smart parking and simultaneous traffic monitoring at crossroads and intersections.

“For such applications that need cameras pole-mounted at significant heights, multi-sensor cameras guarantee high-performance image capture, while reducing the time and cost associated with installation and maintenance as well as lowering the amount of cabling required when compared to 3-4 cameras,” Jun added. “Yet most customers can meet their wide-area surveillance requirements cost-effectively by using pan-tilt-zoom (PTZ) cameras for perimeter detection and fisheye cameras for indoor applications.”

How manufacturers plan to meet the demands

In terms of meeting the AI requirements, van de Looveren said that Bosch’s video management software BVMS 10.0 is already Person Identification ready. Customers can make use of additional licenses offering them person identification based on Artificial Intelligence. Overcoming typical challenges of regular facial recognition solutions like occlusions, various angles, and poor light conditions.

Jun explained that the version 3.0 of IDIS’ Deep Learning Engine is now powering IDIS Deep Learning Analytics (IDLA) and gives customers advanced analytics for up to 80 channels via an optional service module within IDIS Solution Suite VMS. With higher accuracy, this is dramatically increasing control room efficiency, reducing false alarms and enabling a faster, more appropriate response to events and incidents.

What do startups in the field see?

Startups in any field provide a fresh perspective on market demands. In the video surveillance segment, many new startups focus on cloud-and AI-based solutions. Qumulex is one such company. Buckley said Qumulex was started to fill a need that we saw in the market in easily migrating video surveillance and access control customers to the cloud, with four key differentiators. These are unified-video surveillance & access control, hybrid solutions to offer a mix of cloud and on-premise, mobile accessibility, and an open architecture that works with other manufacturers.