Twenty minutes after clocking in to work, the average security guy, who sits in a control room watching a bank of television screens, will be giving no more than 30% of his attention to the video streams. His severely diminished attention will be spread across 8 or more such video streams - that is, unless he’s brought his own portable TV to work to watch, or he’s playing cards with the other security guy. That’s often the way it is.
The simple fact is that the weakest element in the security systems built from large populations of intelligently placed cameras and banks of DVRs is the human watcher. Even if this system component is conscientious, its performance is pathetic. The human is at his or her best when there’s a real incentive to watch the screen. Video surveillance does an effective job in casinos, where the watcher’s attention is drawn to follow specific individuals who (horror of horrors) appear to be winning. (If you like attention then go to a casino and start winning - you’ll get buckets of attention.) In most other situations, the human component needs to be upgraded, because it is simply not up to the job.
What job? - the job of scanning hours of video in order to notice if anything significant is happening.
The truth about video security systems is that they are genuinely useful in three ways:
- The presence of cameras acts as a deterrent to would be intruders.
- The video record will provide useful information in the event of any actual incident.
- Having security guards in the building watching videos can be very useful if a genuine intrusion is detected.
All of which is fine, and may be well worth the cost, but more is possible - a great deal more.
Avocado Security: The Technology
Avocado Security has built a software platform that makes a big difference to this situation (both product and company are called Avocado Security.) The best way for me to describe it is to call it a video mining product. You could say it was a physical security product, but in truth its a video mining product and physical security happens to be its current primary area of application. As far as I know it’s the only product of its kind.
There are two important aspects to the product:
1. Data analysis/video mining: Avocado analyzes video streams. It analyzes these streams to extract a whole series of “attributes” it identifies human beings (distinguishing them from animals) captures their direction of movement, counts them, identifies them (in some circumstances) takes note of background variances (due to weather), spots small changes of detail and so on. In total a video stream is analyzed for 35 attributes and this data is recorded as data.
The data can then be analyzed in various ways. In particular, for security applications, you can compare activity by time of day, or day of week or any other time frame to highlight irregularities. What Avocado Security does, in effect, is reduce video to sets of data that can be analyzed and compared. If you want, it can do this in “real-time” - there’s a lag between an action being captured on video and it being translated into analyzable form, but it’s only about 10 seconds.
This makes a big difference to the surveillance situation because the system can warn someone that something unusual is happening - and it can be investigated immediately - by someone who will give it 100% attention.
2. Avocado Security is a Platform. It is a platform in the sense that you can plug a whole series of components into it. On one side, the components you plug in are surveillance technology - cameras and DVRs - video feeds plug directly into the platform. On the other side you can plug BI tools into the platform. The platform has its own BI capabilities in the sense that it has a series of dials, bar charts and graphics that it has built to highlight specific situations. However as Avocado Security has interfaces to many of the commonly used BI and reporting tools, they can also be used to analyze data in any way that you choose to look at it.
Video Mining
You could say that video mining already exists to some extent in the sense that there are ways to analyze video and extract data from it. You could use face recognition, for example, to identify all scenes from “Raiders of the Lost Ark” in which Harrison Ford appears. And that’s very useful if you’re in the process of cataloging movies. But Avocado Security is doing something quite different. It is caturing and codifying activity.
Some of its customers have already realized that the usefulness of this extends way beyond security. In retail, for example, you may already have cameras deployed that capture the movement of people through a shop, in order to spot shoplifting activity. With Avocado Security you may well be better able to do this because there may be characteristic patterns in shoplifter behavior. But you can also observe the order in which people browse through the store. You may discover, for example, that even though a given product sells well, it is rarely browsed - so you are giving it too much shelf space. You may discover that certain types of customers (men over forty, say) never buy any of the shirts on display, and so on. You cannot gather such data from video by having people watch the video. You need to condense the information into an analyzable form and them point the number crunching software at it.
The exciting things about video mining in this way is that it’s pretty much a green field. We simply have no idea what we do not know about the behavior of people (or animals or cars or truck or whatever). Just as there has been a big pay-off from the data mining of transactional data, there is going to be a big pay-off from the data mining of video streams. It will surely make a huge impact in security and I have little doubt that it will have an impact in retail and transport. But it’s quite possible that it will make a difference in many other activities - in the health sector, in education, in leisure. It’s hard to know, because right now very little of this data has been analyzed in this way.
Avocado Security is breaking new ground. It’s a company to watch and this is technology to watch. You have not heard the last of it.
Subscribe to HaveMacWillBlog in a reader
your “truth about video security systems” is simply wrong. long-term studies have shown that
(a) video surveillance does generally not act as a deterrent;
(b) the records are useless if you want to solve the crime;
(c) the recordings are subject to regular abuse by the persons watching.
all this can be read from the various studies that have been conducted on the massive use of CCTV in england; also, other studies have shown the same results. there might be local variations and exceptions in isolated use, but in general, your “truth” are not.
let me give you a couple of scientific ressources to show that i didn’t just invent that for some reason:
Welsh & Farrington: Crime prevention e?ects of closed circuit television: a systematic review, 2002
Sir Simon Davies, Enquiry into digital images as evidence, 1997
one good article about the uselessness in solving crimes:
http://www.thisislondon.co.uk/news/article-23412867-details/Tens+of+thousands+of+CCTV+cameras%2C+yet+80%25+of+crime+unsolved/article.do
and an overview of “doesn’t work”-articles:
http://www.notbored.org/cameras-not-effective.html
Let’s take this piece by piece.
(a) video surveillance does generally not act as a deterrent;
Your link http://www.notbored.org/cameras-not-effective.html contradicts this directly by maintaining that cameras only deter some people. The suggestion is quite absurd of course, for an obvious counter-example cameras deter speeding highly effectively.
(b) the records are useless if you want to solve the crime;
This is another clearly incorrect assertion that only requires a single counter example to refute. Aside from speed cameras we have the impact that the cameras around London (put there to identify IRA car bombs) had on car thieves. The cameras recognized license numbers and checked them against cars reported as stolen. Car thieves soon learned not to drive into London. (Would that be deterrence by the way?) There is also the thousands of examples of robbers robbing banks and/or shops and getting “caught on camera”. Caught on camera is even a cliché, and you maintain it doesn’t happen?
(c) the recordings are subject to regular abuse by the persons watching
I have no axe to grind here, but this generalization also seems to be a huge stretch to me. Most surveillance systems simply are not linked to video editing suites, so I’m not sure what abuse you’re talking about. In any event I have no axe to grind here. The simple fact is that camera surveillance is a tool and it can be abused - just like a hammer. Regular abuse seems like a huge stretch to me.
Finally..
I find arguments of this kind deeply worrying. There is undoubtedly a Big Brother tendency in the UK and the US that needs to be opposed. Those who argue that cameras are ineffective when they obviously are not, are doing a terrible disservice to civil liberties.
Personally I oppose the general surveillance of citizens because it is indeed effective, not because it is ineffective.
obviously, we both won’t budge much. so i’m not sure what the value of a longish discussion would be. also, please understand that, as english is not my native language, some of my arguments may be stated weakly or even offensive, when i try to be as neutral as possible.
so, just a few short remarks:
- the research results regarding surveillance consistently show that there is little predictable result of surveillance. of course, surveillance introduceses changes and has effects. but to say that “The presence of cameras acts as a deterrent to would be intruders” is not only overly simplified, it’s also not backed up by any study i about surveillance i have seen, at least not as an effect that can be contributed to the cameras. for example, the meta-study i references (”systematic review” from 2002) comes to the conclusion that in those cases few where cameras were effective (mostly car parks) the effects could well be attributed to the enhanced illumination of the area necessary for the cameras. i maintain: cameras are not “effective” in the sense of successful or sustainable, they merely have effects.
- as regarding to speeding with cars, this is certainly a very specific situation, differing starkly from all other forms of surveillance. still, there was an example in austria where there was a “section control” installed at a highway where cars are measured not for momentary speed but for the average speed between two measuring point. the result was that after 9 or so month, the number of accidents involving speeding were up at the same level as before. tickets issued for speeding of course were not. people just learned to get around the surveillance. the same goes for many cases of surveillance. one of the results most scientific studies about CCTV in the UK came up with was that thiefes typically adapted.
- as for the abuse, i was clearly too provoking, and at the same time unspecific in my argument. one thing that goes for most surveillance systems is that the people behind the monitors are not subject to surveillance. as a consequence, they can use pretty much all means available to them to make recordings of the video from the surveillance cameras. often this can be done via equipment that is present, in other cases recorders are brought in and connceted to the system. most the material shown in tv shows of the “americas dumbest burglars” or americas dumbest pedestrians” kind (or whatever they are called) is taken from surveillance systems without autorization. this constitutes one more or less systematic mis-use.
the other mis-use comes from the fact that most surveillance systems are operated privately. as such, these companies are under pressure to show their value by creating as many “incidents” that can be resolved as possible. therefore, typically things that most people don’t care about (and some do, but not enough to really do something about it) such as peeing in a park or between cars, throwing away litter and other kinds of “social crime” are suddenly brought to our (and the police’s) attention because of this conflict of interest we have here. i would call this anther form of mis-use.
also, this shows that of course such systems can be very effective, but in most cases, not in the way intended. as numerous studies in the uk show, the widespread CCTV systems have done nothing to enhance security on the streets or even solve most crimes that happen. what they have done is chance the environment we live in in a very subtle and, for many, aggravating way.
ok, even though i planned to write only a few short remarks, this has become somewhat lenghtish. given that you obviously have a stake in such systems, i still believe that we both won’t budge much. otoh, i am very curious about how the discussion will evolve (if it does).
You seem to tie yourself in knots. Suddenly lighting is a deterrent to intruders but cameras are not. In terms of defensible space there are a whole series of things that act as a deterrent to intruders which include; lighting, locks, planting bushes under ground floor windows, having a clearly identified boundary to a property (psychologically defensible space), having a visible (real or fake) burglar alarms on the side of the building, having a dog and of course cameras. All of these have “effects” as you say. The primary effect is to deter.
As for crime itself, the most effective deterrent known is fear of being caught. Cameras heighten that fear in some would be criminals. This is why cameras that are turned off can deter (when first introduced some speed cameras in the UK had no film in them, but still deterred speeding). This is also why most bank robbers now wear masks.
Naturally crime moves to places where surveillance is less. This is hardly new information. “When the cat’s away the mice will play” has been popular wisdom for centuries.
The idea that CCTV may be badly deployed at times seems credible to me. That would be poor use of good technology. In the UK there may indeed be an over-investment in such systems in terms of cost/benefit.
The statement that “the widespread CCTV systems have done nothing to enhance security on the streets or even solve most crimes that happen.” is worth a little dissection:
“solve most crimes that happen” logically means that they do solve some crimes that happen. (Last time I was in the UK I saw footage of a murder in a shopping center where the murderer was clearly identifiable and identified by the CCTV footage)
“nothing to enhance security on the streets” seems quite likely, if crime tends to move elsewhere - to places of lower surveillance. The cameras move the crime around.
You talk about misuse and complain that the people behind the monitors are not subject to surveillance. Isn’t it a little bizarre to make such a statement while you’re claiming that surveillance is ineffective.
The fact that vendors of anything will propagate propaganda is not new information. It’s being happening since Adam was a boy.
You think I have a “stake in surveillance systems”. I don’t. I just write about things and so I wrote about this. But why accuse me of it? Ad hominem comments tend to discredit the writer.
sorry for the “stake in surveillance systems”. i was misinterpreting.
Leave A Reply