Behavioral Recognition Systems, Inc. will be demonstrating its AISight Cognitive Video Analytics software at the 2009 ASIS Seminar and Exhibits in Anaheim. Eric Eaton, the chief technology officer of BRS Labs recently sat down with us to discuss the technology behind this award-winning product and the impact it will have on the industry.
Moran: You have said that your company’s approach to video analytics represented a “fundamental change in the industry away from rules-based systems to a cognitive approach”, could you expand on that?
Eaton: We believe that the cognitive approach opens the possibilities of how video analytics is being used 
today. One of the problems with analytics has been understanding the context of a scene. If the analytic was monitoring a particular camera, defining rules for that camera can be cumbersome because the computer can’t distinguish normal from abnormal activities. AISight being able to do so is the fundamental change.
Moran: What are the key advantages of cognitive video analysis?
Eaton: The AISight uses a three-stage reasoning process of observe, learn, and respond. The observation stage is used to develop hypothetical concepts, or “hypocepts” in that second phase. This is the learning stage in which hypocepts create a context of the area. The final stage is the respond, in which these learned hypocepts are compared to fresh observations.
It is in comparing these observations against the hypocepts that a spectrum of interpretations ranging from purple which is normal for the scene through blue, and green for substantially abnormal observations. Red is green is the final spectrum that indicates events that have never been seen.
Moran: How long does it take to develop the hypocepts for a particular situation?
Eaton: The baseline set is usually completed in 3-4 days. Over the next 4-6 weeks the algorithms will get progressively tighter.
Moran: Are there instances where a rules-based analytic is more appropriate?
Eaton: There are applications where a rules-based system is more appropriate because it is simpler and more effective. In a situation where you always have traffic moving from left to right and what you need is essentially the answer to a yes or no question—is anything moving the wrong way?—then a rule-based system is the appropriate solution.
Moran: In terms of training the workforce, how prepared are end-users to take full advantage of these systems?
Eaton: The output of the system is very straight forward. When AISight determines behavior is in a certain color band it is e-mailed or stored. Monitors receive a picture of the scene for review and are able to click a 5-second video clip. So there isn’t a high level of sophistication required to understand the output of the system. It’s easy to determine from the still and video what unusual activity has prompted the alert. 
Moran: What factors are driving the adoption rate of cognitive technology?
Eaton: Cognitive video analytics is still early adopter technology. Our market penetration is primarily with forward-looking companies and those that are attracted to the technology due to specific threats in their environment. We have installations in maritime environments, classified facilities, restricted areas, municipal building, and hotels. Some of these have changed their security protocols based on what they have learned about their environment by using cognitive video analytics.
Moran: Will this technology grow beyond clients with enterprise-class architecture?
Eaton: AISight was designed to scale. Its main benefit is in applying it to large networks without having to program every camera so the technology lends itself to large applications. The business aspect is important as well. Large clients want to work with favored integrators and contractors to deploy a system of this complexity and sophistication. We understand this and are forming relationships with integrators so that the product can move through traditional channels into the market.
Moran: What hurdles are you dealing with in cognitive video analytics?
Eaton: Computer vision as an academic study, machine learning as academic study; we continue to need refinement and improvement in that area. There is a lot of great research going on in colleges and universities around the world. That knowledge is helping and it will be translated into products that will be developed by companies like BRS Labs and others.
Moran: At what point does this become more of an issue of training systems rather than people, and speak to the issue of where you see us in the convergence of physical/IT security?
Eaton: Low tech solutions to physical security have traditionally been considered to be guards, gates, and guns. IT security developed apart from this. But sophisticated technology is creating a convergence. Oftentimes IT is at the table and they have a great understanding of servers and software and how to integrate these things. But they may not understand as much about video streams or integrating systems that combine those streams and other data.
There is a demand for training of IT professionals. And as well for physical security personnel who want to have a better understanding of these systems. We’re seeing a lot of that training being done by integrators. We’re seeing a new class of experts that are good at streaming video, server configuration, and physical security considerations. This new class of experts is really a profession that we see continuing for at least several years.
I’ve seen physical security professionals that are moving into IT and loving it. And I see IT people moving more toward security and developing their awareness of those considerations.
I think it is slower with IT because there is already a demand for IT to do other things. They seem to be hesitant to specialize in a particular area because they may not feel that they would be as marketable.
This creates an opportunity a lot of physical security professionals to get involved in technology where their expertise is really needed. And a lot of them are taking this opportunity.
Moran: Is the security industry evolving to open systems and away from proprietary and, if so, where are we today in that curve?
Eaton: A lot of good things have been happening just in the last 12 months. Historically proprietary was even used as a sales feature. If you wanted to be sure no one ever got a copy of your video then having a format that others couldn’t interpret was helpful in that.
That value of that argument is losing quickly as security systems become more sophisticated and one vendor does not have a complete solution. Instead you have integrators that have put together several “best of breed” offerings and so interoperability becomes key.
I think the market is recognizing this and end-users are beginning to demand more interoperability and open standards. There are several movements in that area that are encouraging.
There is a lot of innovation going on in security. Necessity is the mother of invention and the type of threat we face in the world today; the individual or small unit, the terrorist threat is going to have to be met with a wide degree of innovation and ways of stopping that threat. That means a lot of vendors are going to have to work together because you need that innovation to ultimately subvert these threats.
Buying a system that is closed and won’t talk to anybody else will paint you in a corner and keep you from getting to the innovative technology that you need to protect your environment.
To ASIS members I would say continue to insist on open standards. I see a lot of great trends in that area I see RFPs being written with interoperability requirements. So keep pushing. The vendors are listening, the integrators are listening. I have recently seen a case in which an end-user selected one product over another because of interoperability. I applaud that and I think it is what is necessary for the industry to evolve in the next 5-10 years.
BRS Labs
2100 West Loop South, 9th Floor
Houston, TX 77027
Phone: 713-590-5163
2009 ASIS International Seminar and Exhibit: Booth 601
Tags: IT/Logical, New Products // Add Comment »
