How To

How to Build a Surveillance Camera System

Tips straight from the CSO

By Anonymous

December 01, 2006 — CSO — STEP 1: Determine your cameras' raison d'être.

From this you'll be able to decide just about everything else. In my organization, we concluded that

surveillance was primarily for forensic use, which meant it had to:

• Let us rapidly review detailed footage so that after incidents (accident, crime,

  terrorism, overcrowding and so on), we could back everyone up from the event to the edge of our

  property to see who they talked to, what car they drove, where they parked, what bus or taxi they got

  out of, what condition the facility was in at the time (icy sidewalk, wet floor, tools in the area and so

  on).



• Be on demand, because we can't watch the cameras all the time. We wanted to

  be able to call up the camera with the best coverage to monitor situations if and when we were called

  by someone with an ­incident report.



• Allow for targeted "smart" cameras for high-risk areas. We wanted the cameras

  to be able to alarm us when something began to happen.

STEP 2: Take an inventory.

If you have cameras already, survey what departments own them, what they are for, how they are

monitored and what format they use, then decide if any of these cameras can be folded into your new

system.

STEP 3: Get smart on camera technologies.

Determine the effect of new technologies on operations, training, maintenance and staffing. Determine

how clear you need the picture to be (as opposed to how clear you want it to be). This will have an

impact on costs. Here are some of your choices:

• Frame rate. The more frames per second (fps) you choose, the more

  network capacity each camera will require and the more data storage you'll need. A Fortune 50

  company I contacted is using 3 fps to watch cash registers. Its studies showed that no human can trick

  the camera in a third of a second. We are using 10 fps because our analytics software requires a

  minimum of 7 fps and we can use only multiples of five. Note here that if you are expecting to catch a

  baseball in flight you'll want a high frame rate, possibly 30 fps or full motion video, but that's going to

  be more expensive to operate.



• Resolution. Similar to frame rate, but this is not a motion issue; rather, it's a

  clarity issue. The clearer your picture, the more pixels it will have, the larger the file will be and the

  more bandwidth it will consume. We determined that Quarter CIF (Common Intermediate Format,

  352x288, which is the default frame rate for DVR systems and is one-quarter as clear as your TV at

  home) is perfectly OK to provide an image to the police that they can use to find and apprehend a

  suspect. If you expect to get mug shot quality out of your video surveillance program then you'll want

  full CIF (TV quality), but you'll pay for it in terms of bandwidth and storage.



• Compression. There are several standard commercial compression algorithms,

  but the most common, for all the right reasons, is MPEG-4. We went with that because it is economical

  and has all the clarity and ease of use we need. You can add a commercial compression package to

  your MPEG-4 to further reduce your storage needs.



• Storage duration. Thirty days of storage is the unofficial national standard.

  This gives most of us enough time, without overdoing it, to learn of an incident before the data is lost.

  When you get an event of interest, you should plan on dropping that file onto a DVD and storing it

  indefinitelyor at least until all criminal and civil proceedings are concluded, including appeals.

  You can also send events of interest to a separate server for indefinite storage.



• Storage size. I did the math: A single digital camera running 24 hours at just 5

  fps, using MPEG-4 at Quarter CIF resolution, creates 11GB of information. That's 330 gigs per month

  for just one camera! A 500-camera system, not at all uncommon for corporate buildings and campuses,

  will generate 165 terabytes every month. So it's easy to see why keeping pedestrian images beyond 30

  days is economically foolhardy. And remember, this is with cameras running on spare frame rates and

  resolutions. Full motion video (30 fps) will create 990 terabytes, almost a petabyte, of data. If you also

  go to full CIF, multiply that by four, and you're talking a few petabytes a month! Stay on top of the

  contractors and consultants so that they stick to the original engineering discussions, or they could

  easily overdo your system. Mission creep can be expensive. And remember, it's not just storage but

  also bandwidth you have to consider. How are you going to move 5,500GB of digital video through your

  network every month?



• Digital vs. analog. At some point you'll want to convert your images to digital

  to take advantage of its easier storage and better retrieve and search capabilities. But analog images

  can travel farther from the camera, so cable runs will impact your choices. A consultant can advise you

  on the best route to go.



• Pan Tilt Zoom (PTZ). There are two ways to do this. The old-fashioned way,

  which still works really well, is to physically move the camera and change its focus whenever you want

  to get a close-up. Dome cameras have a dark bubble over the camera to mask where it is pointed at

  any given time. Make sure each PTZ has a default "park" position that gives you an optimal view and

  focal length. It is easy to forget to do this and leave a PTZ where it was last focused; this will likely not

  give you the best coverage when nothing special is happening. The other (more costly) way to do PTZ is

  digitally, within the image itself. This will require megapixel images. Zooming in on, say, a license

  plate, will fuzz out the image illegibly if you don't start with sufficient pixel density in your camera

  platform.

• Email
• Print
• Comments
• Digg This
• SlashDot

• Case Study: The ROI of Digital Video Surveillance
• 4 CCTV Lessons from the U.K.
• The Hidden Camera