Intelligent Video has a Bright Future

Oct. 27, 2008
The costs of video analytics are coming down, and the benefits are expanding far beyond security

Video analytics has been in the high end market of security projects for several years now. Airports, prisons, military bases, and other large scale facilities have deployed intelligent video with some success over the years. However, as with any new technology, it is taking a while for video analytics to really make its way down to the small and mid-sized projects. It's not “mainstream” yet, which of course begs the question: “How long will it be before video analytics becomes a staple technology?”

Your answer largely depends on who you ask. Some industry insiders will tell you that video analytics is already mainstream because it's already popping up in everyday places such as banks and grocery stores. Others argue that the technology still needs to mature and prices have to come way down before it can turn a profit for most security dealers.

So where is the happy medium? The truth is that the video analytics game is changing at an accelerating pace. Manufacturers are entering the mainstream market with simplified solutions tailored to fit what the small or mid-sized customer needs. No, the mainstream market isn't anywhere close to being saturated; and yes, security dealers should still be cautious about implementing intelligent video solutions as data still comes in regarding the returns on investment. But the trend is undeniable: The solutions are becoming better, the prices are coming down, and video analytics is beginning its move into the mainstream.

“We're really at the very beginning of the crest of the wave,” says Mariann McDonagh, VP of global marketing, Verint. “It's probably going to take the next 3 years or so before video analytics becomes a widely accepted, mainstream video application.”
So what will this “wave” look like? How will video analytics be priced for the mainstream market and what real-world problems will it aim to solve? These are the questions addressed in this article.


When looking at the critical i nfrastructure market of video analytics, an important function is “intrusion detection.” Large facilities, such as airports, have such expansive perimeters that it's not feasible for a person (or even many people) to monitor the entire perimeter. This is where a camera with video analytics can do a good job of alerting security personnel when a vehicle or person is exiting or entering an area at a time or place that could be suspicious.

However, this is not what video analytics will look like for the mainstream market. In fact, physical security will be just one of many video analytics applications.
”If you look at the mainstream markets in video, the number one and number two purchasers of CCTV equipment are retailers and banking,” says Alan J. Lipton, chief technology officer, ObjectVideo.

In order to provide a variety of useful solutions for retail stores and banks, video analytics will have to be integrated with other systems. For instance, banks could be very interested in knowing when somebody walks up to an ATM but doesn't start a transaction. The camera (with analytics) can tell that a person has walked up to the ATM, but it can't know whether or not a real transaction is taking place unless it has data from the ATM.

“One of the hot applications for us in the retail space is something called return fraud analysis,” says Vidient's McDonagh. “85% of the employee fraud that occurs within the big box retail chain is actually perpetuated by this one type of fraud: Returns that are rung up when there is no customer present in the aisle. So what our particular analytics solution does through its integration with the point of sales system is it knows when there is a return transaction that is logged and there is no customer physically in the aisle.”

In grocery stores, a video analytic that could be handy is “slip and fall.” If the camera sees a person slip and fall in the store, the proper personnel could be alerted immediately to check out the situation.

There also could be energy saving applications for video analytics in the future. For instance, a camera could detect that a room is empty and thus turn off the lights.

And of course, video analytics will aim to give users the ability to search vast amounts of video for specific actions or objects after the fact. An example would be if you had reason to believe that a suspicious red van had been in your parking lot at any time over the past 3 months. Rather than having somebody watch three months of video, you could have the analytics search the video for red objects that are the approximate size of the van (give or take). This is an area of analytics in which IBM wants to excel. The company continues to add to its toolkit of server-based video analytic solutions, building upon its long-standing expertise in databases.
Smart surveillance solutions with finite parsing of video will be key moving forward, says Bob Angell, senior informatician, retail emerging business opportunities, IBM. “We can put the events in a database of structured content,” he adds.



“Most of the money in video analytics is coming from critical infrastructure—airports, train stations, subway stations, electrical grids, water treatment plants, petrochemical plants, and throw in a few corporate headquarters. That's where the dollars are,” says Steve Goldberg, CEO, Vidient. “Do we think that there's an opportunity in small/medium enterprise and then even moving past that to consumers? Absolutely. That's the Holy Grail for Vidient. That's what our investors are looking for—the mass market.”

It's no big secret that the reason video analytics is not in every bank, store, restaurant, and school today is because the return on investment isn't there yet. But the key word here is “yet.” Things are changing; and for more reasons than meet the eye, prices are coming down.
Steve Birkmeier, director of marketing, Arteco IVS, says that for clients looking to get into basic video analytics, the cost could be about $1,000 per channel. Arteco also has a new dealer program designed to help dealers maximize their return on investment when it comes to the intelligent video market.
Garry Clarke, president, ioimage LLC, agrees that analytics can be as low as $700 to $1,000 per channel. “We're getting to the point where video analytics will go down to the small integrator and the small installer,” he says.

One way that analytics can be used more efficiently and thus drive the price down is with software that can schedule multiple cameras through a single analytics channel, explains Mulli Diamant, vice president of sales for On-Net Surveillance Systems (OnSSI), which has such software running on its network video recorders. Diamant gives the example of a video surveillance system that has 50 cameras, but not all of the cameras need to have analytics all of the time. Depending on the environment and risk factors, those 50 cameras could be scheduled to run through one of the ten analytics channels at different times. This way all 50 cameras are capable of having analytics, but you're only paying for 10 analytics channels.
Not only does pricing vary among the manufacturers, but so does the manner in which pricing is done. Some companies charge per channel, whereas others might charge per algorithm or behavior.
SightLogix is pricing video analytics in yet a different way. “The real measure of the cost of analytics should be ‘reliable coverage area per dollar,'” says John Romanowich, CEO, SightLogix. “The end-customer's true cost includes costs of everything from cameras to communication equipment to bandwidth QoS and integration to make it work with other systems, infrastructure costs (e.g. specialized structures such as stabilized mounts) and finally t he cost of managing false and nuisance alarms.”

“The most significant trend or innovation in video analytics is really about ‘platform' more than anything else,” says Alan J. Lipton, chief technology officer, ObjectVideo. “What is really unique is this ability to take analytics and really put it into any single device in the video ecosystem and be able to do some cool stuff with that. So now it's in your camera. It's in your router. It's in your encoder. It's in your DVR or NVR or wherever you need it to be.”

Texas Instruments is one of the many company's with which ObjectVideo is partnering. TI's Da Vinci technology-based DSP's are bringing more capabilities on-chip for surveillance cameras, and thus reducing costs. One of the capabilities that can be programmed is video analytics. Yvonne Cager, video security marketing manager, Texas Instruments, sees this as another way analytics is coming down in price and heading into the mainstream market.


No matter how great any video a nalytical algorithm is, it can only be as good as the image it's analyzing. These algorithms are essentially analyzing groups of pixels, and if the pixels comprise a distorted image then the analytics won't be very useful.

When indoor cameras are being tried outdoors for analytics without success, it should come as no surprise that if you're not using the camera for its proper application, then the image will suffer and so will the analytics. But getting a great image isn't the end of the story. These images still need to be analyzed.

The next question is where will that computer processing take place? It's commonly done on a server. However, getting all that video from the camera to the server eats up a lot of bandwidth which has its own costs.

Increasingly, video analytics processing is becoming more distributed. Rather than doing all of the processing on the server, the analytics are starting to be done at the camera, the NVR, etc. This can decrease network traffic without sacrificing intelligent video.

“Agent Vi is not using the standard approach,” says Gadi Talmon, vice president of business development, Agent Vi. The company's video analytics solution uses a new architecture, called IPoIP (Image Processing over Internet Protocol). It's a distributed approach with some of the analytics done locally at the video encoder and some of it done at the server. With this architecture, Agent Vi aims to retain the strengths of local and server architectures, while minimizing their weaknesses.

In summary, video analytics are coming to a store (or bank or school) near you. It's not a matter of “if,” but “when.” Anywhere you see surveillance cameras today is an opportunity for video analytics tomorrow. Most of the applications for video analytics will be outside the realm of physical security, but security dealers who understand the technology will be in a prime position to get the business.