Just where do video analytics stand, really?

 - 
12/09/2009
Unisys is one of the more active companies in trying to disseminate information about security to the world at large, conducting surveys regularly and reporting the results and often offering commentary on emerging trends. And you know they're all over the Twitter. Their PR gal hits me up all the time, but in a fairly passive way - just seeing if I need a voice... Anyway, their latest predictive piece for 2010 is headlined "Unisys Predicts 2010 Yields a Biometrics Boom While Organizations Go on the Offensive to Protect Data," which is all well and good, but it doesn't get at the piece of the prediction that got my attention. Here's the bit I found most interesting:
7. Smart surveillance – Surveillance systems will be become more sophisticated and intelligent. Unisys experts say that real time event detection technology will soon be able to identify a security breach as it occurs and initiate an action instead of simply recording footage to be reviewed after the incident. Improved digital camera technology coupled with intelligent software enables surveillance footage to be combined with other available information, such as facial recognition data, to create alerts so that immediate appropriate action can be taken. Surveillance software will also soon be able to recognize recurring patterns, or individuals to detect when an unusual event is occurring in real-time.
Um, huh? "real time event detection technology will soon be able to identify a security breach as it occurs and initiate an action instead of simply recording footage to be reviewed after the incident"? I was sort of under the impression that we could do that five years ago. I mean, wrong-way detection is fairly commonly used at airports at this point and is exactly what they're talking about - a breach happens (someone goes the wrong way) and an action is initiated (an alarm goes off on someone's desktop and security is alerted). Is the analytics hype so great that what the analytics companies say they could do a while ago, and is supposed to actually be in practice right now, Unisys, a pretty big security integrator by any standard, doesn't even think they can do yet? But maybe they can do it soon? Am I reading too much into that? Maybe. But, to reiterate: "Surveillance software will also soon be able to recognize recurring patterns, or individuals, to detect when an unusual event is occurring in real-time." Soon. Of course, BRS Labs say they can do this right now, and it's not like they've been wall-flowers about it. Does Unisys not believe them? Has it gotten to the point that cynicism regarding intelligent video is so pervasive that the coin has flipped? That analytics companies can now actually do more than most people think they can?

Comments

I think Unisys should BRS Labs :)

[...] This post was mentioned on Twitter by Sam_Pfeifle, Security Feeds - RA. Security Feeds - RA said: SSN: Just where do video analytics stand, really? http://bit.ly/5o9GhC [...]

Wait, John, is the missing word in that sentence "buy" or "sue"?

Bottom line, analytics will never work in the way consumers expect from having seen analytics in action on TV, until and unless we have a revolutionary leap in the way processors work. No way can anyone make a computer perform usefull analysis, including judgement calls, learning behaviours, and filtering false positives, with the level of computing technology we have available at the end of 2009.

Sure, there are any number of suckers actively seeking such a system, and any number of con artists and flim flam men itching to take their money (although I will be overly generous and point out that most of the con artists probably believe most of their own hype). But artificial intelligence (and that's what analytics boils down to, isn't it?) is simply a pipe dream with today's level of technology.

What will happen next week is anyone's guess, of course, and technological breakthroughs happen all the time. But basing your entire plan on the hope that someone you don't know will invent the technology necessary to make it work is a terrible idea- just ask Al Gore and everyone at the Copenhagen Climate Conference.

I'm not nearly as pessimistic as CameraMan. I think there are a number of applications where analytics work - as I indicated with the directional analytics, for example, I've seen them working at Tel Aviv airport in person and hear about success with that application a lot. Many successful analytics might be more accurately described as just really well refined and intelligent motion detection, but there are a number of companies that can reliably identify a person and ignore tree branches, water, etc., and their technology is great for protecting perimeters where motion detection alone creates crazy false alarms.

Judgment calls and learning behaviors are another matter, but I don't think it's outrageous to say that video is reliably identifying potential threats and alerting the appropriate people of those threats. That's why I was surprised by Unisys' predictions.

From what I've heard, UNISYS installed BRSLab's AISight in a major Bank in Brazil to monitor ATM lobbies. I think UNISYS is referring to BRS in this statement..

Hi Itstrue (feel free to identify yourself),
I don't really have any way to verify that, but the question would remain: "Why is Unisys using the future tense?" Especially if they've already installed it and it works. Wouldn't they be touting their ability to do it right now?

Yes, video analytics has its uses as Sam mentioned. As long as the "creators of hype" do not try pushing it for video verification of burglar alarm events, then we should all get along fine. Video verification is tough enough for human operators - a computer has no chance.