Image cropping, people counting, object tracking, tampering detection and exposure modification of portions of a video picture also are possible. Some surveillance cameras include metadata about what is shown in the video, provide the time and date when the video was recorded, and perform face finding.
Of course, how smart a camera has to be is a function of each individual application. Dumbing down cameras is one way to reduce the cost of an installation, unless the alternative is more expensive, points out Fredrik Nilsson, general manager of Axis Communications AB, Chelmsford, Mass.
He cites the example of a location where a central server is required to analyze the output of every 10 cameras, but another 2,000 cameras are to be added, which would necessitate the addition of 200 servers.
“It’s getting expensive and impractical to have that on a central location, so it is better to have it in the cameras,” Nilsson declares of analytic intelligence.
Some examples of cameras that contain their own “intelligence” can be found in products from IQinVision, SightLogix, Bosch Security Systems Inc., Axis Communications AB, JVC Professional Products Co., VideoIQ Inc., and ioimage LLC, to name just a few. Professionals at these firms describe some of the intelligent camera features available and how they work.
FACE FINDING AND WATERMARKING
“Face finding is the first stage of facial recognition,” explains Paul Bodell, chief marketing officer for IQinVision Inc., San Juan Capistrano, Calif. “It’s a level above motion detection.”
In face finding, the camera selects something it thinks has a high probability of being a face. “It doesn’t compare one face to another, but says, ‘We see something that looks like a face,’” Bodell continues.
The company’s intelligent dome can do digital watermarking in the camera.
“Traditionally, that’s always been done in the DVR, but the problem with the DVR is you can feed bogus video into a DVR and have it authenticated,” Bodell maintains. “But if you do it in camera, everything that comes out of the camera will be authenticated. It adds a whole new level of authenticity to the video.”
DIRECTING LEGACY CAMERAS
An intelligent camera system designed for outdoor perimeter surveillance instructs legacy dome cameras on following objects of interest.
Cameras from SightLogix, Princeton, N.J., have extensive video analytics built in them for their specialized application, declares Larry Barfield, SightLogix’s vice president of government programs.
“We are by proxy creating an intelligent system that leverages legacy equipment by being interoperable, and we can make use of devices and bring intelligence to them from the system’s architecture,” Barfield explains.
“Once we detect a target, we take that target’s latitude and longitude, convert the data and shoot it to a dome camera and have the dome autotrack the target,” he continues. “Without any operator intervention, we’ve automated the interrogation process.”
Having intelligence in the camera enables the system to be set up quickly and not rely heavily on network resources. “You no longer rely on the quality of the network for your solution,” Barfield points out.
RETROFITTING ANALYTICS
Putting analytics on a single camera of an installed video surveillance system can be done by using a camera with the analytics inside, points out Dr. Bob Banerjee, product marketing manager for IP video products at Bosch Security Systems Inc., Fairport, N.Y.
He cites as an example a high school that is satisfied with its legacy video surveillance system but would like the front door camera to have video analytics. Replacing the analog camera there with an IP one with video analytics inside could be done easily. The output then would be transmitted on a local area network (LAN) to a decoder and into a DVR or — if the IP camera has hybrid coaxial and IP outputs — over a coax cable to the DVR.
Banerjee estimates the cost of an intelligent IP camera with video analytics inside compared with an analog camera with the necessary encoder to be 10 percent less expensive.
ADDING EARS AND AUTOTRACKING
Video is thought of as adding eyes to security, but ears also can be added with microphones, which is a capability of camera intelligence that sometimes is forgotten, Nilsson points out.
“Either recorded or live audio can be used for intelligence as well as to detect in the most advanced way certain words being spoken in an area,” Nilsson observes. “If you have a gunshot or people yelling in a city environment, that attracts the attention of the operator.”
With an audio detection alarm, sounds above a certain decibel level will mark video, initiate recording at a higher frame rate or resolution, or alert an operator.
Some cameras have built-in microphones or inputs for separate ones. With a separate microphone, a camera can be mounted high, but the microphone can be mounted lower so speech can be heard and recorded.
Putting autotracking within a camera enclosure can speed reaction time as a pan-tilt-zoom (PTZ) dome camera follows a subject, points out Geoff Anderson, national marketing manager for security products for JVC Professional Products Co., Wayne, N.J.
He estimates a 110- to 150-millisecond delay is typical if a camera has to transmit data back through an encoder to a recorder to analyze the image and decide whether an object needs to be tracked. That delay is avoided when the auto-tracking intelligence is inside the camera.
INFRASTRUCTURE SAVINGS
Another advantage to intelligent cameras is savings in network infrastructure. With some that include storage along with video analytics, those savings can be substantial.
“Suddenly it becomes a more cost-effective way to deploy surveillance,” declares Scott Schnell, president and CEO of VideoIQ Inc., Bedford, Mass., especially for installations of only a few cameras.
“Large customers are often the conglomeration of many small sites,” Schnell points out. “Some multinational corporations have lots of locations and need a cost-effective deployment model for putting four cameras on this site, and six on that one and 20 on another.”
With analytics and hard-drive storage in each camera, the most common surveillance problems, like a hard-drive failure, is limited to one camera, Schnell points out.
IP cameras without intelligence and regular network video recorders cost about the same as VideoIQ’s cameras with the video analytics and storage inside, Schnell maintains.
Compared with standard IP cameras streaming video constantly on a network, Schnell calculates his system is 20 percent less expensive because his can run on an existing IP network, since the cameras only transmit video during alarms.
Savings become even more substantial in small installations, Schnell maintains. “Because we can scale down to three, five or two cameras, that opens up whole new markets where it was never economical to put in the basic infrastructure to support a surveillance system,” he says.
Intelligent video cameras can be used as sensors, points out John Whiteman, vice president and general manager for North America of ioimage LLC, Denton, Texas.
“You can go into any type of application today and turn a camera into a proactive sensor,” he notes. “What that has done is to begin to broaden the market for video analysis from high-end nuclear power plants into a broader base of schools, universities, border crossings and certain infrastructure. Car dealers are another application. We do hundreds of those around the country.”
Who has a bachelor’s degree, a master’s or a Ph.D. in camera intelligence? That’s for you to decide and it may vary depending on what your customer needs. But to be most cost-effective with your installation, be sure to employ cameras only at the intelligence level really needed.
