I had the privilege of meeting Dr. Bob Banerjee, senior director of training and development for NICE Systems, for the first time at ISC West 2012. During our meeting, he kindly explained physical security information management (PSIM) to me in language I could actually comprehend. At the end of our discussion, he encouraged me to contact him with any technology questions I might have.
I took Banerjee up on his offer, and we recently discussed the effectiveness of outdoor perimeter protection. He told me that due to a high number of false alarms, most end users deactivate their video analytic-based perimeter intrusion systems within days of going live. It got me to wondering, how can electronic security contractors install an effective perimeter protection solution? In the following Q&A, Banerjee offers a few tips to help integrators deploy a proficient system for end users.
Why do you think outdoor perimeter protection is not as successful as it should be?
Bob Banerjee: When setting up a detector — it doesn’t matter what kind it is — there is usually a sensitivity level on it. For example, let’s say you’re pointing at a tree and you wanted to take somebody that’s walking in front it. What you’re looking for is movement, but you don’t really care about the leaves just gently swaying in the wind. So, you decrease the sensitivity to only capture the really big things that happen, like when a bad guy walks through a scene. However, if you eliminate the small things, there is a chance that you will miss the bad guy coming in, too. If there is a risk that you will miss the bad guy, then you change the sensitivity so much that you basically capture everything because you’re frightened that something is going to be missed. The problem is the level of sensitivity is such that you’re capturing many kinds of movement, not just the one you’re interested in and that’s what generates all the false alarms.
Would it be effective to install a thermal camera around the perimeter?
Banerjee: Thermal cameras are brilliant because they are detecting heat. However, the cameras can’t see faces or clothes; they can only see a blob of a human walking around. What they can do very well is they detect that there is a person in a place where there shouldn’t be one at a distance in horrible weather.
The problem is with thermal cameras is that they cost between $10,000 and $20,000 at the minimum. Unless the end user is a very high-end facility, they’re not going to have hundreds of these around the perimeter fence. Also, because you can only detect that there is someone bad, you cannot use it for identification. You have to couple it with normal cameras as well.
What types of cameras would you suggest integrators install?
Banerjee: Most people put fixed cameras along their perimeter fence because they can record everything. Then, there are two options for deploying video analytics. Integrators can either install a smart camera, in which case it has the brains to detect that someone’s doing something bad in front of them. That’s intelligence at the edge. Or, installers can have cameras bring it all back to a recorder with the intelligence to look at the video and see if someone is climbing over a fence or whatnot. That’s centralized intelligence.
Fixed cameras are great because they look over an area, but the problem is they cannot zoom in and look at the fine detail of a person or situation. So, with a fixed camera looking over 200 yards of fence line, if Bob Banerjee climbs over the fence of this particular point, all users see is this little black blob climbing over a fence. They won’t be able to see any detail. A number of companies are pairing pan/tilt/zoom (p/t/z) cameras with fixed cameras. A fixed camera detects that something is going wrong and it tells the p/t/z camera to zoom in on that situation. The p/t/z camera can zoom in like a telescope and give very fine detail of just that particular area. That’s the best way to do perimeter intrusion detection.
How can integrators determine the amount of cameras to install along the perimeter?
Banerjee: Assuming that an integrator is working with a straight, long fence, an approximate rule of thumb is to install a fixed camera every 100 to 200 yards. Of course, the shorter the distance, the more accurate because the larger the person will appear. Because p/t/z cameras offer 20x to 30x zoom, those can be spread further apart.
Is there anything else that you would recommend to integrators?
Banerjee: My biggest issue is the reason why video content analysis has not been popular in the past. Salespeople will show integrators a 10-second clip of a bad guy being caught but then they don’t show the 20 clips that they have of exactly the same thing but where there were false alarms. This has been the big problem with using video analytics for perimeter intrusion detection for the last decade.
I would encourage integrators to demand from vendors recorded evidence — at least 15 minutes — of a camera in various lighting and weather conditions proving that there are no false alarms.