It’s the difference between today and one day.

From simple, been-there-seen-it functions manufactured into or enabled by cameras to encoders and then to complex video management platforms and server-based intelligent storage and retrieval gear, security video analytics has that infamous biometrics baggage blues: a combination of overhyped marketing messages, blended with always remembered failures, mixed with a taste of scientific breakthroughs and stirred generously with a heck of a lot of CIS Crime Scene Investigation fictional gadgets.

During a PSA Security Network roundtable on the topic, integrator Bill Bozeman kicked things off by plainly stating that “video analytics is making a big splash within the security industry but there seems to be a lot of misinformation.”

No kidding.

Stripped of its various names and many definitions, it’s a software programming tool to help organizations maintain their level of security in public and private spaces by automatically detecting and alerting to incidents or potential threats. And it is more, too, as the tool’s abilities spread into the business intelligence side of organizations, especially retail, homeland security and traffic management.
In an effort to explore where analytics is now, where it is going and what benefits it provides, a survey of enterprise security executives, industry research firms, integrators and industry leaders found that there’s plenty of current applications but that the technology is evolving as it overcomes its checkered past.

“Another advantage of business intelligence analytics is that systems not one hundred percent accurate still provide value,” continues Jenkins. “In security applications, end users cannot afford to miss a potential security breach; however, business intelligence algorithms can provide reliable trend information despite not always counting every person that enters a retail or commercial location. It is fair to say that in the short to medium term, security applications will remain the bread and butter.”

Concerning that bread and butter, video analytics sits on two shelves: pixel scrutiny and pattern or object recognition. On one hand is motion detection, around for many years, and camera tampering alerts. On the other hand, objects – from people and vehicles to baggage – and patterns can be “recognized” with the ability applied to license plate recognition, bag left behind, people counting, vehicle counting, and trip wire or so-called virtual fences, as examples.

Whatever the shelf buyers pick from, in its technology heart beats an algorithm. In computer chips and systems, an algorithm is basically logic written in software in order for a person to – ahem – do something. That something starts with an alarm in real-time monitoring or a bookmark to an event when retrieving and reviewing video. When it comes to pixel and pattern/object analysis, algorithms move from fairly simple to very complex, and from an embedded chip to sophisticated software embedded in command and control.

In the tricky world of outdoor surveillance, where seasons change and leaves blow, video analytics-enabled cameras are now available specifically for outdoor use that have substantially more processing power within the camera to filter out and compensate for outdoor variables, enabling more accurate detection over even greater areas.

Archana Rao of research firm Frost & Sullivan believes, “Complex technology combined with the relatively slow migration to IP in the video surveillance space and an uncertain economic environment is hindering the large-scale adoption of analytics. The current video analytics solutions continue to suffer from the unrealistic expectations set in the early stages of the market. Additionally, low awareness about its real benefits, the lack of open platforms and price sensitivity restrain its widespread adoption.”

Still, taking careful steps toward IP cameras and then analytics can yield bottom line benefits. For instance, St. Cloud, Fla., has deployed network cameras from Axis Communications at its water and wastewater plants as a first step in modernizing its existing analog video surveillance capabilities. And future on-camera analytics capabilities with third-party developers also allowed for a significant reduction in the number of cameras required at these facilities, according to Bud Peck, CAD/GIS analyst for the city.

One example comes from the very tough streets of Karachi. The Pakistani government has a $151.7 million “Safe City Karachi” project that is expected to be completed in two years to fight terrorism and crime, improve traffic monitoring and overall public safety and security through video surveillance. The major components of the surveillance system include video cameras that provide feed from all the surveillance points, underlying fiber and other network and back-up on wireless connectivity to securely carry video feeds, the command and control centers to store and monitor these incoming feeds, and video analytics to automate the monitoring and surveillance functions such as automatic number plate recognition.

But, back in the real world, assumptions that the technology reduces manpower, increases officer effectiveness or decreases the need of more cameras or other types of electronic security systems are, well, not necessarily so. What analytics does provide is more information, often at the right time, so that a person can decide to do that “something.” When all is said and done, that’s a pretty good security thing. And in the face of false alarms, some systems now can, on some level, even learn to improve their accuracy.

Like any tech tool, it all comes at a cost, both in dollars and in operational requirements. The approach has to be properly designed, and sometimes this means more cameras and possibly better cameras, more lights, more processing power and more infrastructure, all depending on the application. There must be quality video images coming out of the cameras since analytics depends on the details, suggests Fredrik Nilsson, general manager of Axis Communications. Camera positioning is another key aspect sometimes forgotten in retrofits: angles and views needed in video analytics are often very different than those needed in surveillance situations.

Overall, video analytics should never be the only tool. Take a virtual perimeter. You really need multiple technologies, not just analytics, advises Chris Johnston of Bosch Security Systems. He sees growth within the retail sector, for example with people counting and traffic patterns. For example, a supermarket store manager may know just when to add checkers upfront if cameras can alert on the shopping cart line. There is also the forensics capability, so you can take a recorded scene and then run scenarios.

According to Francis D’Addario, emeritus faculty, strategic influence and protection innovation, with the Security Executive Council, the council is sponsoring a project entitled “Store of the Future 2020” that will integrate video and audio data intelligence capabilities for people, asset, and transaction protection. Current development partners include the Aronson Security Group (ASG), the FireKing Security Group and Verint. “Our initial proof point includes a demonstration of the concept at ASG’s ‘Great Conversation,’ a multi-sector security summit in Seattle this month. Test beds will evolve capabilities throughout 2010 and beyond,” D’Addario says.

The project prioritizes physical and fiscal people risk mitigation. “Associates, partners, service providers, and consumers will benefit from life safety, property crime, fraud and mission critical risk detection,” he notes. “Objectives have been qualified to what the council has identified as board-level risk. The capability to corroborate and report exceptional risk events while enabling key business processes accountabilities including cash control, and sales or transaction oversight (credit, debit and inventory) and throughput are mission enablers.”

Digging into the details of the project, D’Addario adds that capabilities will include video and audio confirmation of burglary, robbery, and theft with exception reporting integration. DVR and NVR capabilities allow instantaneous data packet communication for confirmation of pre-determined events to end users, security or monitoring services. “Package left behind, man down, people counting and retail traffic analysis capabilities may be further leveraged to determine pre- and post-event conditions, and persons of interest,” he says. Elsewhere is this article is D’Addario’s wish list for future tech features.

He admits that “certainly, the range of improvements for video and audio analytics to advance the core mission, protect personnel, assets and throughput is relatively underdeveloped (at this time). The capabilities to enhance the consumer experience, check crime, contain cost, and enhance brand reputation with thoughtful protection of privacy remains relatively un-mined.”

But he feels integrated analytics will exceed expectations. “We will merely require the engagement of cross-functional leadership to understand the requirements of risk-based mitigation as a mission enabler.”

A kind of analytics can take on after-hours patrolling through a camera. At Arizona’s Grand Canyon West Resort, a tourist destination owned and operated by the Hualapai tribe, security video bandwidth and storage requirements are reduced by the use of a feature called ACF – activity controlled frame rate. ACF is built into an IndigoVision IP camera and makes it act like an alarm motion detector. When no motion is detected in the scene, the video is transmitted at a very low frame rate. When motion is detected, video is automatically transmitted at the maximum frame rate configured.

In another application, “improving video analytics capabilities was a key factor in deploying a complete integrated IP video solution,” says Rich Tieslau, the Marietta, Ga., director of management information systems. “It took too long to find an event with any of the old video systems.”

Some providers now have a module menu of specific applications. For example, March Networks’ includes a family of intelligent video analytics, which include left object detection, loitering detection, slip and fall detection, and others. Case in point: a large banking customer in Europe selected a loitering detection analytic to alert staff to anyone loitering in an ATM vestibule beyond a set amount of time. Once they receive an alert, security staff can view the video in real-time to determine whether that person is a potential security threat, or might be involved in ATM skimming (bank card fraud).

Still, the potential of video analytics cannot be denied. Automated sensors never tire, can cover large distances, and “see” what the human eye would miss, even in poor or zero light conditions. People can then make the subjective decisions based on credible information to form an appropriate response. 

Video analytics offers the ability to move the security function from a reactive to a preemptive posture by providing real-time alerting and monitoring as events unfold. Further, the technology is also a foundation to obtain better information for investigative purposes after an event has occurred. Specifically, video analytics that employ GPS-based detection will position a PTZ camera to track, zoom in and archive high-detail information for forensic use. Thus the automation afforded by video analytics can now be used to alert, monitor and ultimately aid prosecution efforts.

It is important to note that not all analytics software is a best fit for all situations. Some are better at bags left behind, others at motion in a dark tunnel. It is crucial that decision-makers and operations center personnel understand what analytics solutions to tap given a specific situation, suggests Dave Fowler of VidSys. Also, video can be driven by a variety of analytics solutions including analytics engines that are correlating events from multiple systems/devices. These all have a role into effective video analytics and data analysis, he says.

Advances continue, nonetheless. Take predicting of human behavior. A consortium of European researchers, coordinated by the Computer Vision Centre of Universitat Autnoma de Barcelona, has developed what it calls Hermes, a cognitive computational system consisting of video cameras and software that the consortium suggests is able to recognize and predict human behavior, as well as describe it in natural language. It’s based on three different focus levels: the individual as a relatively distant object, the individual’s body at medium length so as to be able to analyze body postures, and the individual’s face, which allows a detailed study of facial expressions. The information obtained is processed by computer vision and artificial intelligence algorithms, which permit the system to learn and recognize movement patterns, contend members of the centre.

But, is it really possible to analyze and discover potentially unusual behavior, based on the video information gathered, and send out accurate warnings?

Well, that sounds like situational awareness nirvana or another example of over expectations – what got video analytics into trouble in the first place.

But, then again, the security industry, enterprise security leaders and integrators continue to have a love affair with “gee whiz” stuff, no matter its immediate effectiveness or bottom line business benefits.