Rick Santoro, who’s in charge of security for Trump Entertainment Resorts in Atlantic City, N.J., would seem the ideal candidate for using video content analysis — that is, technology that helps organizations draw intelligence from their surveillance video. Hotel and casino operators like Santoro’s company are known for having cutting-edge surveillance, the better to prevent loss and fraud on their high-stakes gambling floors.
Yet Santoro is taking a cautious approach. He’s only now testing systems from several vendors and integrators that could help his security group monitor places like storage facilities, hotel lobbies, parking garages and event venues for the US$1 billion company. But he doesn’t think analytics tools are good enough to capture, real-time, the kind of sleight-of-hand movements his group is watching for on casino floors. Instead, Santoro is looking to find ROI with basic applications — such as setting up cameras in liquor storage areas so that they record only when motion is detected. With an emerging technology like video analytics, he says, proceeding slowly is smart.
“The longer we wait and research and look, the better the technology that we’re seeing is, and also the cheaper and more reliable it is,” says Santoro, whose full title is executive vice president of asset protection and risk management. “Five years ago was light-years behind the technology that’s around today.”
That’s one statement that everyone seems to agree with. Video analytics has come a long way from the hyped-up, gee-whiz technology of a few years ago that promised way more than it could deliver. “A lot of people had bad experiences, especially with the outdoor [analytics tools],” which were expensive, hard to configure and didn’t always work, says Sandra Jones, principal at an eponymous security consultancy that focuses much of its attention on video surveillance. “The pioneers have a lot of arrows in their back, but I have to give them a lot of credit because they were ahead of the technology curve.”
The good news is, the technology has improved enough that organizations, slowly but surely, are finding that analytics tools can help them make sense of all the video they are collecting and even find an ROI — but only if they are careful shoppers. Here’s what to know before you begin.
Tip 1: Understand the marketplace
Once, video analytics was largely a software business, with applications residing on central servers or digital video recorders. These applications were (and still are) based on algorithms that monitor for specific events — motion detection, intrusion detection, entry through an exit and so on. Some of the analytics companies are still focused on only the software. For instance, the flagship product for Aimetis is supposed to work on any standard, networked PC.
Increasingly, however, the industry is moving toward embedding the software into hardware devices and selling the whole thing as a package, says James McManus, research analyst at IMS Research. Longtime market leader ObjectVideo has stopped selling its software to end users. Instead, the company develops software that runs on digital signal processing (DSP) chips that are manufactured by original equipment manufacturers (OEMs) such as Cisco, EMC and Texas Instruments. The OEMs then install the DSP chips onto back-end storage devices, networks, digital cameras or encoders.
(An encoder is a device that translates analog video feed into digital information. In other words, you don’t need to have digital cameras in place to do video analytics, as long as your analog cameras are compatible with encoders that can translate the signal into a digital feed.)
Some large companies, including Bosch, Honeywell and Sony, are building analytics into the digital cameras they already sell. Other specialty companies such as Verint Systems and NICE Systems focus on analytics but develop and sell the whole kit and caboodle — from video management software, to cameras and encoders, to the analytics software itself. This shifting marketplace leads to some strange bedfellows. ObjectVideo and Verint Systems are competitors but partners as well: Verint has developed some of its own algorithms and also licenses some of ObjectVideo’s algorithms and sells them on its own hardware.
Tip 2: Start with your business need — then select the technology
The latest video analytics tools claim to do very sophisticated activities, from identifying loiterers to detecting vandalism to monitoring crowds for dropped baggage. When evaluating your options, you may be tempted to get carried away. Don’t. Always start with the business need, then see if there is technology that could fill it — not the other way around.
“It’s like every other decision,” Jones says. “What is the return on investment, what is the value it can bring my organization, and what can it help me accomplish that I can’t accomplish any other way?” Could analytics allow you to reduce your security guard force? Could it let you monitor a site remotely and save money on gasoline? Could it help manage all the video information you’re collecting or let you conduct investigations more efficiently? Prices have come down, Jones says, but the technology is still expensive.
One way vendors are dealing with this is by moving to packaged models, with groups of algorithms targeted at specific industries. Cernium, for instance, got its start selling software that allows airports to monitor for people entering through an exit lane. Now the company (which has licensed its software to the OEM Toshiba) sells packages for education, cultural institutions, gaming, governments, retail and other industries. Even within these packages, however, you might not need all the tools.
At Trump, a lot of the technology “is overkill for us,” Santoro says. “We are looking for basic, simplistic ways to alert our people to changes in areas. We’re not inclined to go with too many bells and whistles. A lot of the systems have so many features and so many things that you end up not using them.”
Tip 3: Think about whether on-the-edge analytics makes sense for you
Another key decision is whether you want to have content analysis performed “on the edge” — that is, on digital video cameras or encoders, rather than on servers or DVRs. Analysts say this is where the industry is heading, but right now you still have a lot of options.
The advantage of on-the-edge analytics is that content analysis can be performed when the video is of its highest quality, before it is compressed to be sent over the network and stored. A traditional, centralized model, however, provides more flexibility. One DVR or server can do analytics on more than one camera feed, which means that capabilities can be directed and redirected based on the needs of the minute.
Which direction to go, says Frost & Sullivan research analyst Dilip Sarangan, depends largely on your organization’s network capabilities and what the IT department is willing to put on the network, since video traffic tends to be a bandwidth hog. The decision may influence which vendors you want to consider. Some vendors, such as Cernium, focus on centralized tools, while others such as Ioimage focus on on-the-edge setups. A lot of vendors, however, do sell both.
Tip 4: Test, test, test, before you write the check
Once you identify your business need and narrow down the field, it’s time to start testing. Video analytics technology is able to deliver on more of its promises than it could a few years ago. Even today, however, the technology must be configured correctly, and it may not work at all in certain situations.
“Of course [the vendors are] going to say it’s great; it does all this kind of stuff,” warns IMS Research’s McManus. But be suspicious. Integration is a problem. So are false positives. Fortunately, vendors may be willing to let you try out the hardware or software for a month or two before you actually write any checks. Insist on it.
That’s what Brian Ishikawa at the Bank of Hawaii did. Ishikawa, vice president and director of corporate security at the Honolulu-based bank, was interested in one of the more pie-in-the-sky applications of video analytics — 3VR Security’s digital video recorders, which incorporate facial biometrics. The company claims that its facial recognition technology allows customers to search through video archives and find all the times a given person shows up on camera.
For Ishikawa, the pitch was powerful. If someone cashes a forged check, Ishikawa’s group might be able, without much research time, to look for other instances where that same person had appeared, possibly cashing other forged checks. Not only would this allow the bank to ensure that the same investigator was working both cases, it could also help aggregate small check fraud cases to make prosecution easier.
Ishikawa borrowed a test unit to install in one of the bank’s branches for a trial period. “You can set the system to a percentile scoring of possibilities,” he says. “If you set it on a higher possibility, it’ll give you fewer photos to view. If you lower that percentage, you may have a lot more false positives, but you might capture the party with no glasses and no beard.” Overall, he was happy enough with the results to put together a business case, which persuaded management to invest in some of the devices.
Even now, though, he’s proceeding slowly. The system could, conceivably, be used for marketing purposes, with alerts set up to notify bank employees when an important customer walks in the door. But Ishikawa hasn’t ventured there yet. In fact, he hasn’t even firmed up plans to install the DVRs in all 90 bank facilities, because he doesn’t want to end up with something that’s widely installed but out of date.
“The technology is moving so fast,” he says, “that a lot of times it’s hard to make a decision, enterprisewide.”
Comment: [email protected]
