Finally, one of the panel members, Gadi Lenz, suggested that it might be time for someone to step up and put together a very good, in-depth book that can be used for general industry training on video analytics. Just as there are scores of training books for video surveillance in general (and new books for network video surveillance now on the market, too), it’s time for this info to be available for video analytics.
Get past the architecture debate
One of the dominant discussions in the video analytics field has been the issue of edge architecture versus centralized (server-based) architecture. At the conference, the general consensus was that the best architecture moving forward is a hybrid approach, where edge devices like cameras or encoders create metadata that accompanies a video stream, and that perhaps some of the final processing (generating alerts, using rules engines) is done at a more centralized point. But for now, said Bozeman, the typical integrator believes that the best format for VCA is to put all the intelligence at the edge. There will need to be some further convincing and education of the system integrators of the best models for architectures, and training on the flexibility of different architectures if integrators are expected to get beyond an “edge is best” belief.
Get profitable and consolidate
Integrators pin their business model on working with trusted partners. They may step away from the highest-tech companies if they know the second tier (in terms of advanced technology) is going to get them better service and stand behind their product…and that the companies will be there in the coming years to provide advice and additional support.
With a down economy and many VCA firms running solely on venture capital investments, that makes integrators nervous. Having already seen Steelbox close its doors this year (another company generally thought to have had very good technology, but not enough capital “runway” to get them fully launched and stable), integrators have to be concerned whether today’s partners will be available for them in a year’s time.
Bozeman told the audience that he wants to see VCA companies become profitable, because if they’re profitable, that means they are more likely to be solid partners. There was also a general expectation from our panelists that we would see some consolidation of the small, niche video analytics companies into the big players. We’ve obviously already seen this happen from Honeywell (they acquired ActivEye roughly 2 years ago) and Tyco (they acquired Intellivid last summer), and the expectation is that this will happen with the other “big players”. While many times integrators loathe the day that their favorite, trusted small technology firm become part of a bigger organization because they often lose the personal interaction, the panelists generally felt that associating video analytics with the Honeywell-level types of companies would help build trust. They felt that the end-users/purchasers were more likely to put their trust in this technology if more of the big companies are actively involved in this technology.
Provide the technical data
There was a fair deal of general frustration with the amount of technical data available on the video analytics systems currently available. Integrators on our panel said they need detailed, honest technical information. They generally felt like they could, with a fair bit of training get 80 percent of the way through such systems on their own, but they needed more in-depth the technical specifications to get a job 100% right. In the end, this is part of the greater “education” issue that the video intelligence industry needs to address.
Less false alarms
Companies which install security systems in addition to access control/ID systems and video surveillance have always been very cognizant of the issue of false alarms. On the monitoring side, it can mean unnecessary dispatch. If monitored internally, it can lead to a “boy who cried wolf” scenario for the security department, leading to headaches. At the same time, these integration companies want to make sure they’re not missing alarms. Panel members and audience members were aware that VCA companies have to provide that perfect balance between not generating too many false alarms (if it’s a security application, rather than an operations configuration like people counting) and not generating the alarms that are needed. It was generally accepted by vendors that the more canned video and real-world scenarios that they can test such system on, the better the analysis would be – and presumably that would lead to fewer false alarms.