It’s Time to Reassess Video Analytics
Video analytics has been both one of the most exciting and maligned technologies applied to security applications. Find out why its promise of automatically detecting events and generating alerts based on surveillance images is no longer fraught with shortcomings.
Technology and software applications are like raising children. They both require careful nurturing, care and attention when they are brought into this world. Some basic discipline must be used to ensure they don’t get out of control and tarnish the family name and reputation. Finally, it’s advisable to have enough financial resources to “raise” technology and children to ensure they both become well-adjusted and productive members in society. This brings us to the topic of looking at the growth path and development of video analytics.
I think we are through the early and tumultuous teen phase and beginning the formulation of early, more responsible young adulthood. You know, that phase with raising children when complete strangers say something like, “You have such well-mannered and polite children, you should be proud!” You are immediately beaming with quiet pride and utterly astounded at the same time. Technologists that have championed video content analytics should be proud they have stayed with and helped “raise” this valuable technology tool.
So what makes video content analysis different today and worth reconsideration? As with the computer and network technology, given time, systems improve to mature their bodies (hardware) and their brains (software). Video analytics has done both during the past five years. Along the way, manufacturers of these systems had to reset expectations and improve application training for the systems integrator to prevent “overselling” the solution.
Analytics Undone By Early Antics
When video analytics first came on the scene back in 2004, OK it was earlier, but I mean when it became a marketing darling with great pizzazz and sizzle, the possibilities were exciting to say the least. Video as a security tool was now ready to become an active participant in the war on crime. We finally had a proactive vs. reactive tool for your customers and an exciting new product that would justify purchases of your video systems. When something sounds too good to be true …
I bear witness to this because I had a front-row seat to video analytics. I got a chance to work with a very intelligent and talented group that comprised the brain trust of the former Phillips Electronics. While my role was focusing on how to help with marketing challenges, the “recovering engineer” within me kicked into a full-blown relapse. The possibilities to solve the old security problem of improving response to incidents seemed endless. Video that could be proactively analyzed in real-time and then alert different people in different places within minutes? WOW! This was manna from heaven for security professionals in both the public and private sectors and systems integrators supplying the products.
The possible applications of this new technology were only limited to one’s imagination and therein lies the initial problem. There were rules, application guidelines, and limits to the software’s ability to learn and recognize the content it analyzed. The variables in an outdoor video scene that must be accounted for are nearly limitless. Technology analyzing them was not. The video analytic products were overhyped without enough small print, and they were oversold by eager integrators who had not realized the nuances of the technology.
The result was predictable: disappointed security customers and red-faced systems integrators that vowed to never sell this technology again! This unfortunate series of events can be summarized as follows:
- The initial marketing hype over-promised what the technology could actually and reliably deliver with high confidence factors. When you consider first responders will either act or not act based on a security technology tool, it better be in the mid to high 90% accuracy range. Most of the time, this technology was realistically in the 70%-80% range.
- The systems integrators that were early adopters, yours truly included, were in for a steep learning curve of how environmental and camera lighting/positioning factors would impact performance. This often resulted in “detuning” a system to bare minimal tolerances that reduced false positives, but eliminated the practical use of the system’s true benefits to the customer.
- The third fatality was those early customers that allowed us to experiment with their facilities and careers. They were often disappointed and vocal, rightfully so, about their experiences. As we all know they have very long memories!
3 Platforms Pack Promise
So why should you reconsider video content analytics? First of all, imaging has both improved and dropped in price, which means the software that analyzes the raw data has better quality of information and more of it to work with. Better info in means better results, or for us old-timers “GIGO” (garbage in, garbage out).
Second, the sophistication of software engineering and competition has increased through learning what has and has not worked in the past. Small changes in software algorithms based on field deployments allowed the software engineers to build better and stronger code.
Third, the applications where analytics deliver security and business value have been refined with practical field deployment and customer feedback. Camera placement, scene illumination and field of view best practices provided vital feedback for better performance. Simply put, there were contributing factors that have stalled market adoption.
In my opinion, there are three main deployment strategies and application platforms to consider. Each of these three categories applies to different security challenges based on your market needs, financing and commitment. Let’s look at these in more detail: