Video surveillance turns the corner

Video surveillance has been around since law enforcement agencies started using closed-circuit TV cameras in the early 1960s. Today, the basic principle might be the same, but what goes on behind the camera is a far cry from a security guard watching a monitor. Digital and Internet-based technologies have made it easier and cheaper to set up and maintain surveillance systems. These days, a camera could be watching you on just about any busy street corner.

Many agencies are already using basic systems that let human operators monitor live scenes remotely and store videos for later review. And recent advances in software and hardware promise to further reduce costs and expand coverage while delivering analytical capabilities that can greatly enhance security.

The newest frontier in video surveillance is content analysis software.

Instead of having human operators sit in front of monitors 24 hours a day and try to stay alert, software can now send an alert when a specified activity takes place. For example, a person crossing the U.S./Mexico border or jumping over the fence at a port facility could trigger such an alert.

Some vendors, such as VideoIQ and ObjectVideo, specialize in that technology.

A major component of advanced analytics involves not only recognizing objects but also tracking their movements. TerraSight software from Pyramid Vision, a Sarnoff company that specializes in aerial and tower-mounted surveillance for defense and homeland security, can “fingerprint” an object to continually recognize it as it moves through scenes. TerraSight also filters irrelevant content, such as video artifacts -- Moire patterns or pixellation, for example -- and lighting changes.

“When you’re working with surveillance, you want to know where something is and you want to know right away,” said John Bradburn, Sarnoff’s senior business development manager. Each video pixel in TerraSight can have a geospatial coordinate associated with it, and an extensive metadata catalog helps analysts track events. Sophisticated algorithms move from one camera to the next to ensure continuous views.

Another new technology is adaptive learning surveillance software. BRS Labs developed what it calls a cognitive video analytics system that examines video for behavior patterns. The company’s AISight software tracks objects frame by frame to detect variations from established patterns. When it sees such a variation, it sends an alert.

“Let's say there is a camera monitoring a parking lot, for example. [AISight] is going to learn that cars tend to park in particular discreet locations,” said Eric Eaton, BRS Labs’ chief technology officer. “It's going to learn that people tend to go from a car to a particular entrance or exit point on the scene. And it's going to learn that those patterns of behavior are normal. Then let's say that somebody were to move around from one car to the next car rather than just going to one particular car. The system is going to alert on that behavior because it doesn't match the pattern of behavior that it has learned.”

Play, record, rewind

Of course, the capabilities of new software depend to a significant degree on the capabilities of the hardware on which it runs. Most video surveillance setups involve an array of analog and digital cameras and the hardware to transmit, store, play and analyze video.

Storage is a subdiscipline all its own. In the world of analog, closed-circuit TV systems from vendors such as American Dynamics, Bosch Security Systems, GE Security, Honeywell and Pelco, the key recording and storage technology is the digital video recorder (DVR).

Encoder devices bridge the analog/digital gap, which is important because analog equipment can be too expensive for many organizations to replace all at once. However, even though their images can be converted to digital format, analog cameras lack many of the features of digital cameras.

“It’s going to be a great image, but you’re limited in what you can do digitally,” said Jon Hughes, video product marketing manager at GE Security. “You’re typically not going to be able to, after the fact, zoom in for detail.”

Furthermore, the demands on computing power are one reason why high-end analytical tools are difficult to employ on a broad scale, said Fredrik Nilsson, general manager of the North American division of Axis Communications, a vendor of network video equipment. Although he has a customer that runs analytics for eight cameras on a single server, most machines can’t handle that many, Nilsson said.

Analytical software’s artificial intelligence always requires fine-tuning, although that situation has improved as companies make smarter cameras. “The fact that more and more processing is happening in the camera means that servers can handle more and more cameras at a time,” Nilsson said.

Digital cameras that use IP don’t need — but are often compatible with — stand-alone DVRs and can generally operate with a run-of-the-mill information technology infrastructure, such as storage-area network, network-attached storage and Ethernet Category 5 cable with RJ-45 connectors. In contrast, analog CCTV uses the National Television System Committee (NTSC) broadcast standard via coaxial cable.

Nilsson said a typical Dell server with Axis equipment can store 30 days of video from about 100 cameras; anything beyond that requires dedicated storage.

Digital over IP can reduce costs by consolidating most functions in the server or the camera, which makes the system more scalable. Nilsson said the break-even point is about 40 cameras, even though digital units cost 50 percent to 70 percent more than analog ones. However, IP-based systems can also push the specialized requirements of video management and surveillance to IT departments that probably have little familiarity with them.

But digital video has another characteristic that makes it better for surveillance, Nilsson said. Analog’s NTSC standard employs interlacing that draws the video in alternating lines, while digital cameras employ progressive scans — the top-to-bottom, linear painting of an image familiar to anyone who has watched a DVD movie. When frames are frozen in analog video, artifacts can distort the image. “If you ever want to freeze the frame and watch one frame of the video, you’re going to get a little bit of tearing and reduce your ability to identify the person,” Nilsson said.

It starts with the eye

Any surveillance system starts with the cameras, which must have adequate resolution for human operators or analytical software to pick out vital information, such as license plate numbers. They must also be able to capture the entire scene you want to monitor, either by moving around or by having a stationary wide-angle — and typically high-resolution — view.

There are two main choices: high-resolution, fixed-position digital cameras or pan/tilt/zoom cameras. High-resolution fixed cameras could come to dominate the market because they can do digitally what PTZ cameras can only do by physically moving. According to a recent Frost and Sullivan analysis, even though a well-placed high-resolution digital camera costs more, it can do the job of several PTZ units at a lower overall cost.

Choosing the right lens is a discipline all its own. “You’ve got to make sure you’ve got the right lens with the right camera to get the images you’re looking for,” said Hughes, adding that camera resolutions control network decisions. “If you have a normal MPEG-4 stream coming off a camera — non-megapixel, VGA resolution, 30 frames per second — it’s probably going to be 2 to 3 megabits/sec for that stream,” he said. “If you have 3 megapixels, it could be as high as 40 megabits/sec. You really need to have a large, gigabit-type network to handle multimegapixel cameras, and you need to send that full image across.”

To reduce traffic, you can put a DVR — or the IP equivalent, a network video recorder — on the edge of a network, or you can choose to record only when certain alarms are activated, Hughes said.

Tearing out the wires

Wireless networks are starting to replace wired ones, particularly outdoors. But anyone considering them should be aware of the trade-offs in image quality, which is crucial in analytics, said Josee Desjeans, wireless product manager at Verint Video Intelligence Solutions. “In wireless, there are always constraints about the bandwidth you can use,” Desjeans said. “It’s very important to balance the video quality with the bandwidth that you have.”

One increasingly popular option is wireless mesh networks, which transmit via a decentralized mesh of nodes rather than relying on centralized routers as Wi-Fi does. However, both technologies can work with each other and with wired networks.

“The benefit of mesh networking, in general, is it requires less infrastructure planning,” said Bill Dickerson, founder, president and director of AgileMesh. “The various nodes can be deployed in an ad hoc fashion, and they all find each other.”

Wireless mesh products from AgileMesh, Firetide, Motorola and Tropos Networks are catching on with law enforcement agencies, which use them to quickly set up video surveillance networks at incident scenes or special events. Dickerson said mesh networks provide unique advantages by allowing signals to go around buildings instead of experiencing interference by trying to go through them, as Wi-Fi and cellular signals do. That is especially useful in hostage situations, in which the ability to monitor a scene from all angles is critical but setting up a video network can be difficult or even dangerous.

Some police departments have permanent mesh installations. In 2006, the Los Angeles Police Department deployed Motorola’s Motomesh wireless nodes and cameras and Canopy wireless broadband platform in the Jordan Downs public housing complex in the crime-prone Watts area of the city. The result has been a 19 percent reduction in crime in Jordan Downs and a 39 percent reduction in surrounding areas, said Sgt. Dan Gomez, the officer in charge of LAPD’s Tactical Technology Unit. Police have used the video surveillance to stop crimes in progress and even to prosecute a murder case. And no camera has been vandalized.

However, LAPD learned an important lesson: Don’t overlook the strain video can put on existing networks. The Jordan Downs mesh network is connected to some of the city’s wired networks, including fiber. “That created a challenge we did not anticipate,” Gomez said, adding that officials had to replace some switches so the network could handle the new traffic.

Standards on the horizon

Experts agree that the biggest hardware trend in video surveillance is H.264, a standard for a type of MPEG-4 format that enables more compact transmission and storage of video. In addition, Hughes said, an extension of H.264 and MPEG-4 called scalable video coding can improve video network scalability by varying the frame rate and resolutions of substreams in each video bitstream coming over the wire, depending on quality requirements. For example, a surveillance camera that only needs to capture seven frames per second at low resolution can share a line with a camera that captures high-resolution, full-motion video at 30 frames per second. Another emerging video standard, MPEG-7, allows more metadata tagging of videos, which improves searchability, Hughes said.

On the wireless front, Desjeans said the IEEE 802.11n standard now in draft form will dramatically improve Wi-Fi throughput. Meanwhile, WiMax, a metropolitan-area broadband Wi-Fi alternative that has taken years to catch on, has not been used for surveillance activities because of its single-input, single-output paradigm. However, a multiple-input, multiple-output version of WiMax released two years ago is being put to use, Desjeans said.

One area that cries out for interoperability is the link between network video devices and services, including video-management systems, analytics and PTZ cameras. Nilsson said a group started by Axis, Bosch and Sony, called the Open Network Video Interface Forum, hopes to publish a final standard later this year, when the first products will likely hit the market. He added that a new version of the power-over-Ethernet standard — 802.3at — will become the recommended power source for PTZ cameras, eliminating the need for separate power supplies.

High-definition TV surveillance is also starting to take off, with vendors such as Avigilon specializing in it.

All those technology trends add up to higher resolutions and faster frame rates on cheaper, more portable hardware. And they mean more realistic video and smarter, computer-assisted surveillance.