The evolution of video surveillance has had a great impact on video analytics. In the transition from analog to IP, and then from low to high resolution, data has grown at an overwhelming rate. This in turn has created the need for video analytics to be able to most effectively use and analyze all this unfiltered information. As video standards continue to become more robust, the need for more intelligent analytics also increases.
One of the key issues has become how to use video analytics not just as an effective, forensic tool to analyze relevant data related to a crime that has already taken place, but as a prevention tool as well. Truly intelligent video analytics involves a decision and reaction support system that uses and analyzes data from various processes, applications, sensors and imaging devices. In a bottom line way, this means that a criminal or a terrorist may leave certain clues before an incident, such as a car parked too long in a prohibited area, unattended objects or stolen license plates. Systems that relay information to authorities in real-time can possibly help stop a crime in progress or even before it takes place.
However, as the demand for video analytics increases, so too do questions about specific applications and how best to implement this technology. One of the recurring questions is: 'where is the best place for video analytics: at the edge in the camera, or as software embedded in the servers?' There are also questions about what qualifies as truly intelligent video analytics and open architecture and what makes the most sense in the adoption process-to purchase a system with analytics built in, or to have it added later on as the need arises.
At the edge and embedded in the system
In regards to whether the system should have the analytics at the edge or embedded, integrators need to look at the purpose of the application. For instance, in a situation in which the user needs video analytics to know if a person has trespassed onto the rooftop of a high-security building, it will generally suffice to have a system that can distinguish between a human form and something else, like a bird. If the only need is to know whether it's a person or a bird (or some other kind of animal) on the roof, then having the analytics built in at the edge, right into the cameras, will suffice because there is not a need to cross reference a database to make any sort of identification.
However, extending this same situation a bit, what if security spots a person in the parking garage at an unusually late hour and wants to try to identify the individual via facial recognition? To do so, security personnel may have to cross reference various databases, in which case having a system where the analytics are built into the software of the system will be most beneficial. Other situations where that also makes sense involve container recognition, traffic monitoring and license plate recognition (LPR). In all these applications, the need to share and cross reference information is crucial, so that merely having the analytics at the edge is not sufficient.
The adoption process
There are often questions as well about the adoption process related to video analytics. In some cases, a user may not have a pressing need for video analytics and simply deals with their immediate security requirements, with the idea that down the line if they need analytics they can always bring in another vendor who specializes in that. The problem that many end-users have found out from hard experience is that analytics is not a simple add on and that dealing with a different vendor for that can be hugely disruptive to their whole video surveillance system.