Eye on Video: Examining intelligent video

[Editor's note: Popular SIW columnist Fredrik Nilsson gets back to his roots with a series titled "Eye on Video". In this series over the next year, he'll be addressing such topics as compression technologies, megapixel video, TCP/IP, wireless communications and other topics affecting surveillance systems users and installers who are making the transition to today's digital, networked, IP-based surveillance systems. His column will appear monthly on SecurityInfoWatch.com and in select issues of Security Technology & Design.]

Dec. 2007 -- Over the last year, we wrote a series of articles that offered readers in-depth insights into various video surveillance installations that featured specific benefits of networked video, be it wireless installations, scalable solutions or bandwidth capabilities. In addition, we showcased how network video was used in retail, gaming, and government settings. The network video market has been going through a rapid growth phase over the last few years, and to some extent the growth has been driven by new innovative technologies such as Power over Ethernet, progressive scan, and non-mechanical pan/tilt/zoom. So what other technologies are in the pipeline that might further fuel the growth of the IP market? In 2008, I am going to focus on these future technologies that will drive the network video market; technologies such as megapixel resolution, intelligent video, compression technologies, and new networking technologies.

This market will not continue to expand by looking backwards; all of the companies that manufacture, install and service networked video systems have to continually look forward and become leaders by anticipating customer needs, by taking risks, and by delivering products with new technology that substantially improves the customer's experience. In order to maintain a successful path, you need to continue to anticipate customer needs a few years from now and deploy new technologies. One of those new technologies is intelligent video.

Intelligent video, also referred to as video analytics, is the process of analyzing video data with the goal of transforming it into useful information. One of the greatest value propositions that intelligent video offers is the ability to alert authorities in real time if something suspicious happens. The capabilities and benefits are numerous, but being realistic and setting the right expectations is key. The challenge is that with any new technology, there is always confusing misinformation that leads to incorrect perceptions - the market "myths" so to speak. What are the myths of intelligent video? Let's have a look at the most common, and try to demystify and clarify them.

Myth 1: Intelligent video is more intelligent than you

Intelligent video is based on algorithms put together by engineers and configurations made by the end user, all of which make the video surveillance system do exactly what the engineers and end users tell them to do. In other words, you set the parameters that tell the camera what to monitor and what to ignore. That being said, however, a network camera with video intelligence that is correctly adjusted will be a lot more accurate than humans when looking for certain behaviors. Cameras never take coffee breaks nor do they take naps. Like a computer, intelligent network cameras are very good at doing one specific task over and over again without getting bored just. But remember, it is not smarter than you.

Myth 2: Intelligent video is the wave of the future

The first intelligent video functionality in a network camera, video motion detection, was launched more than five years ago. People counting, license plate recognition, and camera tampering alarm just started being deployed on a wide scale in 2007. Other functionalities such as "slip and fall" detection are still a few years away, while 3D facial recognition might be even further in the future. So depending on what functionality you look at, intelligent video might be the wave of the past, present, or the future.

Myth 3: I've seen the demo, so this must really work!

It is easy to see intelligent video at its best in a small, controlled environment. The question is:Will it work from long distances, outdoors and indoors, and in challenging physical environments with less than ideal settings? A lot of companies are showing successful demonstrations of intelligent video, but before you get too excited, make sure it works in a real environment. Ask for references from companies that have used these applications in larger installations with more than 100 cameras before you sign on the dotted line.

Myth 4: It takes a PhD to set it up

There is no question that some intelligent video functions are harder to set up than others. Configuring a system means striking a balance between not missing essential situations and reducing false triggers. However, there are some functions that are as easy to activate as the switch of a button. With one switch, the camera system will learn the scene, the parameters from which to record, and will provide video based on your specifications. But some advanced analytics might very well require a PhD. If that is the case, please make sure his/her expertise and/or fees are included in the price!

Myth 5: You need a digital signal processor (DSP) to run intelligent video

A DSP is a specialized computer chip for processing large amounts of data. There are also general purpose processors and more specialized ASIC (Application Specific Integrated Circuit) processors. While some intelligent video algorithms that consume a lot of computing power need to run on specialized DSP chips, most intelligent video software can run on any of these platforms. They can also run on central servers or in the network camera. Running them in the camera can dramatically reduce cost and increase scalability, however server-side or recorder-side systems can be applicable when you're dealing with legacy camera installs. A well written intelligent algorithm is efficient and should use as little computer power as possible. Additionally, Moore's Law is working in our favor (Moore's Law is named for Intel co-founder Gordon Moore, who suggested in 1965 that computing power would double every two years).

By demystifying the myths of intelligent video, we can approach the rest of this article series with a basic understanding of what intelligent video is, what functionality is new and not so new, how it works and how well it works, in addition to the power and expertise needed to make it work efficiently. In future articles on SecurityInfoWatch.com, we will discuss different aspects of intelligent video, including architecture, audio intelligence, pixel-based systems and algorithms, object-based systems and algorithms, and facial and number plate recognition to name a few. Those articles will start this coming summer, as we take the new few months to cover such issues as new compression technologies, megapixel video, the transition to IPv6, crash testing and more.

As the need for surveillance increases, we feel these and other cutting edge technologies such as megapixel, networking and compression, will spur additional growth in the market and will help organizations achieve greater operational efficiencies that improve their bottom line while protecting against physical threats and safety concerns that are avoidable with the help of smart security strategies and implementation.