Last month, we talked about how technology has finally advanced to deliver on the promise of video content analysis (VCA), or video analytics, such that it is finally stable and reliable enough to become a viable consideration for video surveillance system design. The past two years’ dramatic advances in hardware, firmware, software, mobile communications and tools for mining unstructured data are enabling security integrators to add value to security and business operations.
All this adds up to 2013 perhaps being a breakout year in the real-world deployment of video analytics. I called two knowledgeable visionaries of video/facial analytics that service two very distinct market segments for some help. We’ll look at how this is developing, including a discussion of system design architectures, applications and the buzzword of the moment, Big Data. My thanks go to Animetrics CEO Paul Schuepp and 3VR CEO Al Shipp, both of whom shared their time and perspectives for this two-part examination of one of the industry’s most exciting and promising technology areas.
Collaboration Is the New Imperative
Shipp and Schuepp both show an open attitude and vision toward how their respective technologies could be used as tools in a bigger toolbox. 3VR utilizes best-of-breed software instruments developed by others in its own product solution offerings, and is willing to share its own tools with others. Animetrics approaches its evolving market with a similar mindset.
This is one reason why video and facial analytics will grow substantially as compared to traditional security technologies. Where proprietary software and limited collaboration has been the rule in the security industry, both Shipp and Schuepp bring a refreshing “what if?” perspective to their company’s technologies and growth plans. They have the right idea.
This will have a meaningful impact in 2013 because this approach unleashes hundreds of knowledgeable and talented companies that attack the problems of security and business in creative ways. Is this realistic to expect in the near future? Ask anyone who bought an iPhone four years ago. The surge in computing power, cloud services and applications that made work and life simpler grew at what rate? Better processing power, better software tools and, most importantly, the school of hard knocks has taught us valuable lessons along the way.
It sounds like we have some options, and are moving in the right direction. The key is to ask the right questions.
Topology Can Take Many Forms
So what is the right approach to processing video analytical data? Should it be out at the edge? Should it be in the middle world of an NVR/appliance nodal location? Should we bring it back to Fort Knox at the enterprise? Or, gulp, should we rely on the new world order of the cloud? Drum roll please … it’s all of the above.
The right answer again lies in where your customer lives today and where they want to live when they retire. It’s about the right migration path for them that have so many different actions, options and realities to consider. If you had a hard time answering this question, then think how your customer is feeling. It’s kind of like going to your favorite restaurant and taking a friend. The menu has 10 pages. You go right to the two or three pages to firm up what you always order anyway, right? How long before your friend is ready to order? Left unassisted it could be days. You helpfully suggest some good items. Do they still fret about a choice? Usually they are so relieved for some clarity due to the time pressure to order, the decision is made. Your role is to help your customer decide.
Comparing the server side video content analytics to edge analytics is like comparing a 300-pound heavyweight prize fighter with a 116-pound bantam weight. One is big and powerful and the other is light and fast. Stored video is typically Big Data any way you slice it, which means large storage and processing power is required. Not a cheap date.
Consider the past limitations facial video analytics encountered by systems integrators. First, the realities of the environment where cameras were installed had a significant impact on the validity of the video data captured for analysis. Other considerations included the angle of the camera, the lighting/shadowing of the face and the “ear-to-ear” orientation of the face to the camera. These were hard-to-control variables that delivered a wide variety of results. This brought about both false positive and false negative results, which did not instill confidence in either systems integrators or end users.
One can argue the merits of processing video content on the edge or back in the server depends on the application, scope of video input, how the analyzed video will be used and, most importantly, how quickly it will be used. Each processing method has pluses and minuses. Time, bandwidth and processing power come to mind. Are we talking about real-time or forensic video analysis? This depends on the customer’s security application, the threat timeline, criticality, usage policy and the environment.
Oh yeah, let’s not forget about their budget! Your first step is asking these questions I just posed, then making sure you understand the answers.
Big Data Can Bring Big Value
So what makes Big Data so BIG? Simple answer, there is a lot of it. The real answer is how data is viewed and managed by IT types. Structured data is data that fits into a well-defined protocol that IT uses to manage business processes or military operations. Unstructured data is like that crazy, evil twin who lives in the closet and nobody knows what to do with them. This scares the heck out of those rational IT types. It is the discipline of giving some structure (analytics) to unstructured (raw video streams) data and then making some sense of it that we all have struggled with during the past five years.
The National Institute of Standards and Technology (NIST) has testing for facial bio recognition in both uncontrolled (unstructured) and controlled (structured) environments based on facial angle and everyone’s nemesis — environmental lighting/shading. In the past we didn’t have the tools to work with unstructured data. That is changing.
As mentioned in last month’s column, untethered, high quality mobile image capture (smartphone) and fast inexpensive transmission of data to cloud-based processing of large data bases is one of the reasons video analytics will start delivering on its real potential. According to 3VR’s Shipp, in the commercial/industrial/retail world, facial and special video analytics have made significant strides in the past three years.
Pioneering new technologies is not for the faint of heart; it takes both determination and financial backing to cross the chasm to mainstream technology tools. It means real-time information from video that can support more informed decisions for security and business operations. Business intelligence mined from real-time video across enterprises allows dramatic and measureable reduction in fraud, as well as delivering insights into marketing and customer service.
Paul Boucherle, Certified Protection Professional (CPP) and Certified Sherpa Coach (CSC), is principal of Canfield, Ohio-based Matterhorn Consulting
(www.matterhornconsulting.com). He has more than 30 years of diverse security and safety industry experience and can be contacted at firstname.lastname@example.org.
Page 1 of 2 pages 1
Video Surveillance · Systems Integration ·
Al Shipp ·
Big Data ·
Convergence Channel ·
Paul Boucherle ·
Commenting is not available in this channel entry.