Project Deployment: How to Quicken the Pace of Analytics
“You first have to know what you’re setting it up for,” points out Dave Fowler, senior vice president of product development and marketing for VidSys Inc., Vienna, Va., which integrates video analytics systems with its software. “It becomes an issue first of what you want to detect, and then selecting the right vendor that has the analytics for those activities.”
David Ngau, product manager for video analytics in the marketing group of Honeywell Security, Louisville, Ky., points out that analytics are not always necessary on all cameras at an installation. For a job with 200 video cameras, Ngau points out, “During the initial discussion phase, you might find analytics are only valuable to 10 of those cameras.”
According to Craig Frost, Honeywell’s director of sales for video analytics, “Analytics is a tool, and in doing an application, it gives you more data to make the correct decision.”
David Marra, chief technology officer and a founder of videoNEXT, Chantilly, Va., emphasizes that the technology “is a force multiplier and an attempt at increasing the camera’s capability.”
Jim Talbot, CEO of Ionit Technologies Inc., Northbrook, Ill., lists several improvements in video analytic systems.
“First, the manufacturers have increased their hardware’s excess capacity to not only comfortably accommodate third-party software but also have capacity to spare,” Talbot asserts. “Secondly, business analytics convergence has taken place, meaning vendor performance claims have decreased and product performance has increased. True vendor capabilities are aligning with customer expectations.”
EASE OF CONFIGURATIONAnother important element of efficient deployment of video analytics is the ease of configuring. Some systems require exact measurements with a tape measure or similar instrument of the real dimensions of a scene they are viewing. Early systems required reconfiguring if the view changed or the camera moved.
Newer video analytics systems are promising easier and quicker set-up, which for security dealers and systems integrators means faster turnaround on jobs, more accurate calculation of the labor costs for configuring and greater profit.
Marra of videoNEXT explains that a number of manufacturers of analytics are developing configuration wizards. “These are a graphical user interface that is intuitive enough to step you through the process and allow the non-video scientist to configure these cameras.”
Some analytic systems have the ability to learn the background of a scene and eliminate alarms for extraneous movement within a few hours or days. “We’re seeing these things get smarter over time,” Marra observes.
Virage, an Autonomy company, based in San Francisco, is creating generic alarms that can be deployed over thousands of cameras in large installations.
“We’re trying to move away from the selection of rules at trip wires or zones, because a lot of the time we really don’t know what the operator requires,” reports Dave Humphrey, Virage’s chief technology officer. “We’re going to allow the flexibility to be able to program a system, but not be programmers. We are not experts in security â€” operators are, and so we don’t want to pin them to a set of rules. We want to give them more flexibility to look for what they want to look for.”
PACKAGED BY VERTICAL MARKETSAn interesting development in the attempt to ease configuration is that vendors are packaging video analytics systems and software by vertical markets, such as retail, financial, transportation and others because of the varying requirements of the analytics for these industries.
“So if I’m a retail operation, I buy the retail package streamlined for the types of things I might look for in a store or warehouse,” Fowler explains.
Marra points out that an analytic program is available for every scenario.
“There is no one universal fit for all applications,” Marra stresses. “Study all the capabilities and pick the best one for the application. It’s a very specific science at this point.”
CONFIGURINGTechnicians should configure as much as they can before they go on site, such as obtaining IP addresses from the customer for the cameras, determining how long after an alert a camera triggers recording, and programming in the e-mail addresses to which the alarms should be sent, Humphrey recommends.
Some companies will analyze proposed camera views they receive from integrators and evaluate how their analytic systems can be configured to meet the customer’s request.
Certain tasks, such as people counting and directional violations, require specific camera angles, Marra states. “If you do a people-counting camera, it has to look down at the door rather than across,” he emphasizes.
One advantage of the straight-down view for counting is that the background does not change, points out Stephen Birkmeier, vice president of Arteco Vision Systems, St. Louis. Customers must know what they want when the camera placements are designed, he says.
Birkmeier’s technicians reported that 80 to 85 percent of the video analytic applications they see are violated area or virtual trip wires.
“The nice thing about those applications is they’re incredibly accurate, they require the least amount of time to set up and they require very little tuning once they’re set,” Birkmeier notes. He estimates experienced technicians can configure a virtual trip wire in 15 to 20 minutes.
Setting up a video analytic system certainly is a learning process, Marra points out. “If it is an exterior analytic, we are going to want to come back and configure the analytics in different scenarios, such as day, night and rain, to get a base line of performance over those different environmental conditions,” he explains.
The seasons can affect calibration of analytic systems, such as whether leaves are on a tree, which introduces extraneous movement that may not be visible in winter when the tree’s branches are bare.
The method of configuring video analytics systems for detecting humans often varies by vendor. “Some video analytics manufacturers try to detect the head and shoulders to detect a person,” Honeywell’s Ngau explains. “Another manufacturer might develop some type of algorithm to detect a person by their relative size and speed.”
Video analytic systems perform better when the object to be detected is larger in the camera’s frame, such as when it takes up 80 percent of the frame instead of 10 percent, points out Scott Schnell, president and CEO of VideoIQ Inc., Bedford, Mass.
CALCULATING COSTSEasier deployment of a video analytic solution may mean increased profit on a job. Integrators can more easily calculate labor if they understand how much time is involved in configuring a system.
Fowler concedes that labor costs for video analytic systems also vary based on an installation’s sophistication and equipment.
“After education of the dealer on camera placement and zones for analytics, the real driving cost will be the installation of the cameras,” Honeywell’s Frost says, adding about analytics, “The more familiar the integrator is with it, the lower the costs.”
The integrators Schnell speaks with tell him they build time in their bids to specially calibrate each camera that uses analytics. “They have also built in some budget for post-installation problem-solving associated with recalibrations,” Schnell reports.
Humphrey advises allowing one hour of configuring per camera â€” easier camera configurations will provide the time needed for more difficult ones.
Marra suggests budgeting approximately four hours for an initial setup and review of an outdoor video analytic system. “It may take you 30 to 40 minutes to go through the wizard, but you need to come back and check it in a different environment,” he explains.
Experience may be the best teacher for cost estimates, Talbot points out. “If you align management and field technicians by having management set up a few systems in the field, realistic time estimates will follow,” Talbot maintains. “Analytics is a service and should be priced as one.”
DONâ€™T SKIP TRAININGGeneral courses that teach video analytics are difficult to find because so many of the systems are proprietary and were developed differently. Vendors and manufacturers offer training specifically for their systems, some of it online.
“The first place we go when we need information is the vendors we are working with,” Fowler declares. “They typically have a training program and best practices, and then we also talk to some of the people who have done installations in the past.”
Humphrey advises dealers to become familiar with video analytics technology. “Don’t be scared of the technology,” Humphrey suggests. “Experiment with it; we always encourage that.”
DIGITAL VIDEO MANAGEMENT SYSTEMSWith the number of cameras and complexity of video systems increasing, Per Hanssen, president and CEO of Salient Systems Corp., Austin, Texas, points out that some video analytics companies are partnering with video management systems like his company’s to simplify their operation and expense.
Other video analytics companies are providing their software as plug-ins. He points out that the movement of video analytics to the edge â€” to encoders and cameras â€” may allow for greater flexibility when combined with additional plug-in analytic programs.
“In the long run, you will see both some manufacturers having their own simple video analytics embedded and an option for plug-in, where the customer selects the video analytics that fits their purpose,” Hanssen predicts.