Scottsdale Healthcare has been upgrading and expanding its video surveillance system with a variety of Panasonic IP and analog PTZ, dome and fixed cameras. Over 125 Panasonic cameras are installed to date with plans to double that number.

More and more surveillance video equipment is using the Internet protocol for communication. To date, IP communications has gained the greatest acceptance in centralized devices, such as digital video recorders. But increasingly, devices located at the network edge, including video cameras, have IP communications capability. Other examples are video encoders that add IP capability to analog cameras.

Although IP cameras and encoders represent a relatively small percentage of the total video camera market today, the situation is changing due to technology advances that are enabling new capabilities and bringing costs more in line with the analog market. In this article we explore these new advances and the impact they are having on system design and economics.
One security company that has adopted IP video in a big way is Vanguard Technologies of Baton Rouge, La. “Eighty percent of what my company does is all IP,” notes Vanguard Technologies President Jerry Jones.
That statement is not surprising, considering that Vanguard has a strong focus on critical infrastructure and commercial applications with multiple locations. In general, the more cameras an installation requires, the more likely designers will specify IP cameras. The reason is that the network video recorders (NVRs) that store video from IP devices generally can handle a greater amount of cameras than traditional DVRs that are designed for use with analog cameras. Even though IP cameras generally cost more than their analog counterparts, the overall system cost may be lower because fewer recording devices are involved.
Vanguard uses IP cameras for another reason, too, though. “Usually if we have a 20- to 50-camera job, we may have 20 percent using intelligence at the edge,” Jones describes.
That intelligence, also known as video analytics, enables a system to generate an alert if, for example, it detects people or vehicles moving in a specific direction. “We also do a lot of fencing — instead of physical barriers, we use video a lot of times,” Jones says.
Traditionally video analytics were handled through centralized processing but several manufacturers have introduced IP cameras or encoders that can provide some video analytic capabilities — and where feasible, Jones prefers that option. The reason relates to the MPEG4 compression standard, which Vanguard appreciates because it conserves bandwidth and minimizes storage requirements.
“With MPEG 4, you do have some artifacting,” explains Jones, referring to the image imperfections that can occur when video is compressed. “You don’t want to post-process it after that. It’s better to have as much as you can as close to the camera as you can get it. It’s more reliable and you get [fewer] false alarms.”
Several technology advances have paved the way for analytics at the edge. “The early days of analytics was very processor-intensive — you needed a very powerful computer to run the algorithm,” explains John Whiteman, president of the Americas for Denton, Texas-based ioimage, a supplier of video analytic systems. “What happened in the last three to five years is that now digital signal processors have enough horsepower to run most algorithms, and some algorithms that started with a server solution are now running on edge devices.”
Depending on the type of analytics involved, it may be possible to eliminate a central processor, which in turn can eliminate a single point of failure. As Henry Kim, senior product manager for video manufacturer Samsung Techwin of Carson, Calif., explains, “If the processor goes down in a central location, the whole system goes down.”
In other video analytic configurations, some processing of the video image is handled at the edge, minimizing the amount of video that is sent to the central processor, thereby enabling the central processor to operate more efficiently. “You used to run into a barrier where most companies could only process eight channels of video per server,” explains Paul Bodell, chief marketing officer for San Juan Capistrano, Calif.-based video supplier IQinVision. “By moving some of the decisions out to the edge, now the server isn’t running full time trying to decode every bit, and the number of channels per server has increased.” Today’s servers, he says, can handle dozens of cameras.

IP Efficiencies

Minimizing the number of servers also saves money in other ways, Whiteman notes. “Initially with multiple servers in a room, you might have 25 to 50 servers for 100 cameras, which generated heat and needed power,” he explains. By reducing the number of servers, system designers can reduce cooling and power requirements.

Not all cameras with analytic capabilities work well with all analytic software, however. Partnerships between camera and software manufacturers have been a key factor in enhancing interoperability. “Every week we’re contacted by more guys who want to start taking advantage of intelligence at the edge,” Bodell comments.

The Open Network Video Interface Forum wants to help make equipment more interoperable in the future. Dr. Bob Banerjee, product marketing manager for Fairport, N.Y.-based Bosch Security Systems, describes the forum’s history. “Axis Communications, Sony and Bosch got together and said, ‘We need to get people to adopt IP video faster, so we’ll come up with a standard way for everyone who writes software to treat all hardware equally,’” he explains.

Some integrators are not letting a lack of standards stop them from pushing video analytics at the edge. Consiliant Technologies, an Irvine, Calif.-based integrator, proposes video analytics embedded in the camera on 50 percent of proposals to its customer base, which includes data centers, schools, police departments and other large enterprises.

“It’s core to our message,” explains Kevin Dailey, vice president of the security group for Consiliant. Salespeople for the company explain to customers that video analytics can mechanize a passive camera so that it can “identify violations and intrusions rather than documenting failures,” Dailey says. Alternatively, customers are told, “they can spend 90 percent of the cost and get 50 percent of the functionality,” comments Dailey. After listening to that pitch, between 10 and 20 percent of customers choose the higher priced but more proactive option.

There are several specific applications where IP cameras with built-in analytics are particularly effective, sources say. Whiteman notes, for example, that an IP camera with internal motion detection may be less costly than an analog camera that uses an external PIR for the same purpose — particularly if the area to be covered is large.

Fredrick Nilsson, general manager of the Americas for IP camera manufacturer Axis Communications of Chelmsford, Mass., adds that in comparison with traditional photoelectric beams, IP cameras with embedded analytics can do a more effective job of counting people coming through a door.

Security professionals need to recognize the limitations of analytics at the edge, however — particularly when no central processing is involved. As Jack Cabasso, managing director of Hauppauge, NY.-based video supplier Aventura Technologies explains, analytic capability often is hard-coded into video cameras, which limits the installer’s ability to adjust the device. “People like to say the cameras are plug-and-play but it usually takes about half a day to install one camera,” Cabasso says, adding that installers often have to return to the site to reposition the camera to avoid false alarms caused by weather or other environmental conditions.

“There are a number of things that don’t require heavy processing and can be done successfully at the edge, such as identifying left or removed objects,” Cabasso concedes. But he argues that, “The real benefit of analytics comes in where you’re doing software functions where you can tweak and optimize it.”

Edge Storage

Another capability that is moving from a centralized location to the network edge is storage. Here, too, advances in technology have made this possible, where it would not have been in the past.

“The capacity of storage that you can put at the edge is increasing dramatically,” Bodell comments. “It’s incredibly inexpensive today to put 32 gigabytes of recording at the edge.”
A video camera equipped with a secure digital high-capacity card, also known as an SD card, can store a couple of days worth of video recorded at 30 frames per second, Nilsson explains. In the event that greater recording time is needed, some video cameras have a connection for a hard drive or other external storage unit.
Another option is to back up the local recording by transmitting video at a lower frame rate to a centralized storage device. Although that video will be of lower quality, the advantage is that much less bandwidth would be required than if the central storage device were recording at full frame rate — and higher-quality images still will be available on the edge device. Reducing network traffic through local storage may be particularly valuable when a costly wide area network connection is involved.
On-board storage also can help minimize the impact of WAN transmission problems. “If you’re going over water or through trees, you may have a bad few minutes or few hours,” Banerjee relates.
Some manufacturers address this situation by using a buffer at the edge that resumes central recording when the network problem is cleared up. Other manufacturers, including Bosch, use what Banerjee calls a “drip feed” approach. “When the network comes back, the NVR says it’s missing one-and-a-half hours and asks the edge to drop feed me while I fill in the missing gap,” he explains.
For applications where it’s suitable to record only in the event of an alarm, integrators may be able to totally eliminate central storage, Bodell advises. “If you have a smaller system — less than 10 cameras with recording that is primarily event-based — then you don’t need an NVR or a DVR,” he says. With today’s local storage options, he says, “there’s enough storage in the camera to save weeks of activity or low-resolution video when there is no event.”
The downside to that approach, however, is that storage at the edge can more easily be stolen, thereby removing valuable evidence from the scene of the crime — another argument for sending at least some video to a central storage device.

Video Management

Some video installations, especially those with a large number of cameras, call for a video management system to simplify the process of retrieving or analyzing video images. When an installation includes video management and local storage, dealers will want to do their homework in selecting equipment that will work well together.

As Bodell explains, “When the NVR first came out, it streamed everything back to one location. But now more sophisticated NVRs and video management software are saying, ‘We don’t care where the video is stored; we’ll collect and present it to you in a nice fashion.’”

Bodell predicts that within the next 18 months, virtually every NVR will offer compatibility with edge storage, which could further reduce the amount of bandwidth that must be transmitted on the network.

Banerjee has a different vision. He believes NVRs could be eliminated by letting cameras handle NVR tasks. “If all you want to do is record and playback video and search, you can put that functionality into the camera and the camera can talk across the network to hard drives,” he says. “The implication is the installer doesn’t have to sell, install or maintain servers, anti-virus software and patches. The camera is a piece of firmware that is rarely updated.”

Some manufacturers already have moved in this direction by enabling individual cameras to talk directly across the network to one another with the goal of simplifying — and enhancing — system control. For example, a fixed camera with video content analysis could be used to detect a suspicious person and to send a message directly to a PTZ camera to track any such people it detects. Banerjee contrasts that with the traditional approach, where the fixed camera would send an alert to the video management system, which would then direct the PTZ camera.

By enabling the cameras to talk directly to one another, Banerjee says, “You bypass the single point of failure in the middle and it becomes a fully intelligent system with no video management system. All you need is a Web browser to look at a camera or a recording.”

Other Trends

While analytics and storage are the most talked-about trends driving intelligence toward the edge of the video network, several others are also worth noting. Fredrick Nilsson, general manager of the Americas, Axis Communications, Chelmsford, Mass., notes, for example, that some IP cameras can accept input and outputs from external sensors. That capability lets the sensors trigger recording — or higher-frame rate recording — under certain conditions, such as when a sensor reveals that the room temperature has reached a pre-set limit.

Another new capability found in some IP cameras is two-way audio. Although eavesdropping laws make audio recording impractical in some parts of the United States, Nilsson says pre-recorded audio warnings are useful deterrents that can be triggered in response to motion detection or another external event.
It’s also worth mentioning that virtually all IP cameras and encoders lend themselves to another capability that can simplify video system installation — power over Ethernet technology, which brings power to a device over the same twisted pair wiring that also carries data in an Ethernet network. As Joe Taylor, chief operating officer, Panasonic North America, Secaucus, N.J., explains, “With the advent of PoE-compliant devices, networked systems allow edge devices to be easily installed and relocated anywhere on the network.”