As more organizations realize the intrinsic value of video in a physical security system, getting more value out of that video is becoming increasingly important. And those same organizations are learning of more affordable ways to record and store surveillance video with much longer retention times. Therefore, the number of video analytic installations is significantly increasing around the globe. SecurityHive reached out to a number of video analytic vendors to discuss the current market conditions and how organizations are turning “alerts and events” into actionable business intelligence.

(Featured image: Intelligent Video Analysis from Bosch alerts operators to potential security risks.)

Thanks to surveillance cameras, GoPros, Dropcams, cell phones, and every other video capture device now available, we’re able to record video at unprecedented degrees. YouTube sees 100 hours of new content uploaded every minute. And there is a lot of information embedded in all those video frames. However, without someone willing to watch all that video, it might as well be uploaded to a black hole.

Enter Video Analytics
According to Wikipedia, video analytics is defined this way: “Video content analysis is the capability of automatically analyzing video to detect and determine temporal events not based on a single image.” This content analysis has become the driving force behind the rising adoption of video analytic packages with their ability to comb through hours and hours of video to pull out what is important to an organization.

Mike Matta, CEO of analytic vendor Solink (http://www.solinkcorp.com) succinctly positioned the underlying corporate usage of video analytics when he stated, “what are the questions you are looking to answer with video analytics?”

Video analytic software capabilities, in direct correlation to computer processing power, have made these event outputs a viable technology. Although there are still so many variables with cameras, camera lenses, color calibration, and video processing that “finding all the woman in blue skirts” is still not a routinely viable process today, Eric Olson, VP of Marketing at PureTech Systems (http://www.puretechsystems.com) told us “Intelligent video is basically the emulation of the human eye and a good portion of our brains. This is no easy task and continues to be an area of continued research and development. While there are emerging solutions to specific problems such as people counting under controlled conditions, contrary to some bold claims of ‘behavioral analytics’ that just ‘learn’ everything, computer vision and pattern recognition, which is the essence of video analytics, will continue to be an area of research and development.”

Wikipedia’s definition aside, what is video analytics in the physical security world? Positioned as specific “behaviors” that can be ascertained within live and recorded surveillance video, these analytic behaviors begin by being able to determine a person from a vehicle from an object. And from these various objects video analytics can determine behaviors such as:

a person moving in area; person crossing a line; crowding; person tailgating; loitering; vehicle moving in area; vehicle crossing a line; stopped vehicle; tailgating vehicle; suspicious object; traffic obstacle; perimeter breaches; people demographics, people travel times, precise people counting, gender, ethnicity, age, dangerous parking, unauthorized removal or theft of objects, as well as the much promoted facial detection, facial recognition (yes, they are different) and license plate recognition (LPR) also known as automatic plate number recognition (ANPR).

Retail is ground zero for video analytics
Retailers and companies targeting their business have been particularly quick catching onto video analytics. Already, they’re using this technology to figure out when stores are the busiest as well as where people are walking, stopping, and looking. Some are even using eye-level cameras to identify items people are looking at on fully stocked shelves. Facial recognition software is also helping stores assess shoppers’ age, gender, and race to target ads and provide accurate data about consumer demographics.

Steve Russell, founder and CEO of Prism Skylabs stated the goal is partially to give brick-and-mortar retailers the type of information that e-retailers already get about what people look at but don’t buy. Having that type of information can help retailers get a better sense of what inventory they should carry even where they should put it in the store.

From a business perspective, these tasks can help with controlling check-out lines in stores, assist with parking management in busy lots, and determine the success of various merchandising or marketing efforts within a retail store.

Next Questions
Solink’s Matta positioned video analytic usage this way: “how to use the analytics you are collecting to determine your company’s next investment.” Matta explained that when a company is looking at a capital expenditure within their production/operations area, or their personnel / human resources (HR) ranks, or even their next marketing campaign, how can you bring the value of what is actually happening in your environment on a daily basis into that investment equation? Can you determine what trends are happening within your company or is something just a fad. This ‘trend versus fad’ aspect also brings into play the value of the amount of recorded video available for analysis.

Looking specifically at video analytic solution providers (see list of vendors at: http://www.securityhive.com/business-directory/dir-category/video-analytics/) some developers offer very specific offerings while others provide a wide range of behavioral analytics.

Elke Oberg, Marketing Manager at Cognitec Systems (http://www.cognitec.com) stated the importance of focusing on a specific behavior –such as Cognitec’s facial recognition FaceVACS-VideoScan solution. “The best technology on the market that can detect and identify persons of interest in real time while computing demographic and behavioral data. It can support security staff, marketing departments, and operations management with one application.”

The expected trend for personal and cyber security is that more and more organizations will deploy facial recognition technology to safeguard our accounts and credit cards as we get ever more connected to the wired and wireless networks.

Facial recognition has the potential for major improvements in the commercial and government sectors, including law enforcement, health care and medicine, banking, computer industry, and even the media and entertainment industry. But facial recognition is not the only use in video analytic systems. So what does an organization need to understand in order to choose the most appropriate video analytic package for their needs?

According to Zvika Ashani, CTO of Agent Video Intelligence (Agent Vi – http://www.agentvi.com) “Usually there is a very good correlation between the size of the installation and the requirement for video analytics as the end users realize that without video analytics it will be very difficult for them to manage their surveillance system and maximize its value. Video analytics helps them get a much higher ROI as its cost relative to the entire project is usually small.” Remembering Matta’s statement, “what are the questions you are looking to answer with video analytics?”

Are you looking for a straightforward facial recognition system or are you trying to protect a critical infrastructure compound with perimeter detection, people loitering, and vehicle parking management capabilities? Steve Tangert, Associate Product Marketing Manager for Bosch Security Systems (http://www.boschsecurity.us) acknowledged “video analytics provide sense and structure to surveillance images and can help alert operators to perimeter breaches, people loitering near a sensitive area, illegal or dangerous parking, theft or unauthorized removal of objects, and much more.”

Real Time Versus Recorded Video
Organizations need to determine if their needs deal with “real time” live events or are they more interested in determining the business intelligence value of “recorded” video. Analytics such as license plate recognition (LPR/ANPR) can fit into real time needs as well as historical analysis. A critical product determination to look for in a video analytic solution is the software’s ability to process video faster than “real time.” In other words, if your organization wants to determine how many customers made repeat visits to your store over the past six months, the video analytic software needs to be capable of processing six months of video in a much shorter period of time. There are packages on the market that can only process video at “real time” speed and in the above example would require six months to give you back that analysis. Not practical.

Bosch’s Tangert added “Live analysis can alert operators to potential threats before they become actual security breaches, or they can alert them to security events as they are happening to improve response time by personnel. In addition, live analysis can help reduce the network load if only alarm video is sent over the network.”

He continued, “Recorded analysis helps in quickly locating video of events that have already happened. Analytics store metadata with video images as they are recorded. The metadata contains details on all objects within, entering, or leaving the monitored areas. This enables operators to conduct forensic searches to quickly locate video of an incident, such as a specific car entering the parking lot or a person traveling against the flow of general foot traffic to enter an exit-only door. Forensic searches don’t require analysis rules to be created prior to the event happening; as long as the analytics are turned on, a rule can be created and searched against in the recorded video.”

Solink’s Matta stated that video analytics really is more about “contextual analysis.” What is happening on the video screen may not in and of itself be enough to understand the overarching business issue at hand. Solink’s solutions are specifically geared around “connecting to other non-video data systems within the enterprise” to add more intelligence around what is happening in the surveillance system. For example, Matta explained about a Solink customer that provides payday loans. Being able to reach into Point Of Sale (POS) systems and tying the fact that in the video no customer is present when the loan is executed would set off a flag to let management know about this unusual event. “Customer-less refunds” are another event that occurs when a POS system registers a refund yet the video clearly shows only the cashier making the transaction.

“Business intelligence is a very relevant application for video analytics and there is growing demand for it. Agent Vi already works with systems integrators to deliver business intelligence applications to a wide range of end users who gain insights into behavior in their retail facilities and other sites,” explained AgentVi’s Ashani. “From real-time event detection, to forensic (recorded) search to business intelligence applications” video analytic solutions are evolving into more than just object detection software.

One of the advantages of recorded video is the ability to re-process it for analytic behaviors not originally contemplated for which the camera(s) was deployed. Some may look at this as function creep and this is another area where privacy issues come into play.

Speed and Power
With the increased computing power that is now being found in typical IP-based surveillance cameras, many video analytic vendors such as Bosch and AgentVi have their software embedded into the camera to provide “edge analytic” capabilities. This camera-and-analytic combination showcases the adoption of these various behavioral analytics systems where organizations are looking for tighter integration and quicker business intelligence dissemination. With camera integration such as this, cameras with built in relays can trip door locks without the need for any other equipment such as a PC in the loop. The possibilities that video analytics in real-time events brings to operators can help prevent potential threats from becoming actual security breaches.

Looking at recorded video Bosch’s Tangert replied “forensic search is an investigative tool to help operators search hours or days of video within minutes to find and collect video evidence.” And PureTech’s Olson added “for a vast majority of institutions looking for video protection, live and recorded video are both required.”

Olson went on to say “video is the epitome of ‘Big Data’ that is packed full of valuable hidden information. Video is the media that is permeating every aspect of our lives. Video is a vast array of continuous data. When you start applying intelligence to the data, pulling in data from related sensors and establishing correlations, the amount of information that can be gained is limited only by the computing power of today’s processors and our imaginations. Critical business areas [that video analytics can augment] such as loss and fraud detection and prevention, merchandizing, store layout and design, shopper route patterns, targeted advertising, style trends, conversion rates, customer experience, employee performance, and safety compliance” can be significantly improved.

As Solink’s Matta highlights, it is how “the organization’s security team interacts with the rest of the stakeholders in the enterprise.” There is the physical security responsibilities of the company but how does this tie into the loss prevention team; the fraud investigation team; the HR department. “Facial recognition and LPR analytics helps us understand ‘how we move around’ and other data within the enterprise will help us decided what we do with this knowledge, thus the creation of actionable business intelligence.”

“Investigators need to be data scientists being able to decipher (video) data that is normal from the abnormal,” Matta continued. These investigators should be able to look for patterns and trends so their company can better protect itself from security threats, look where to best invest in marketing new campaigns, or ways to improve productivity to increase the company’s bottom line.

SecurityHive posed the question “when do you see video analytics becoming a ‘requirement’ for video surveillance solutions” and Bosch’s Tangert replied “video analytics is growing in adoption as improvements in its accuracy have made it a more sought after feature of surveillance systems. Many organizations today consider analytics to be a force multiplier.” PureTech’s Olson added, “The trend toward increased demand for video analytics is analogous to what occurred with anti-lock brakes on cars. Initially they were an option but once the benefits were understood and the technology perfected, they became a standard on all vehicles.”

With the processing power of computers and the video resolution of the cameras continually increasing and improving, it is a logical conclusion that it will not be long before video analytics –with varying levels of behavioral analysis– will be common place in physical security solutions. Although we (editors at SecurityHive) are not convinced that being able to “find all the woman in blue skirts” is truly viable today, we agree with PureTech’s Olson that the continuing research and development will make this a simple task in the near future. Then it comes back to Solink’s Matta “what are the questions you are looking to answer?” What value is there in knowing all these women in blue skirts and when does this video analytic output become an input into another business intelligence system.

Business Impact
It is our opinion that video analytics is the one area of the physical security market that has the power to transform these solutions from “slip and fall” detection systems to integrated business intelligence systems that deliver extremely important actionable data that can be turned into institutional knowledge, which can drive the bottom line of any company.

I look forward to looking back at this article in five years to see what business processes are being driven directly by video analytics. Add your comments below and let us know your thoughts.