Video analytics is one of those hot technologies that gets people excited. The ability to use artificial intelligence-based technology to "watch" video, extract useful information and create alerts holds much promise for security and surveillance applications. In fact, the technology can appear so cutting-edge that it often sounds more like science fiction than reality. And, unfortunately, the excitement around the technology has led to a marketplace where the line between fiction and reality has blurred. For instance, one common myth about video analytics is that it can spot a terrorist in a stadium full of people -- something even the human eye can't detect. Or it can automatically "see" a person cheating at a black jack table, while highly trained surveillance personnel cannot. This article covers the five biggest myths of video analytics technology and practice, and examines the true state-of-the-art.

Myth One: Video analytics can replace or surpass human performance. When confronted with an unlikely claim about video analytics, a good rule of thumb is to ask the following question: could a person watching the video perform the same task? If the answer is no, then it is unlikely that video analytics can do it either. Even if the answer is yes, it is still possible that the abilities of video analytics will be stretched -- remember, humans have been interpreting visual images for about 5 million years. Computers have only been at it for about 40 years.

The Truth: In most operational environments, good analytics software can be configured to be as capable as a human observer at detecting important events such as cars parking illegally, people climbing over fences, people entering restricted areas, and so on. This event discrimination ability makes the technology a significant improvement over other sensing technologies such as buried cables, microwave detectors, and taut-wire fencing. The real value of video analytics, though, is that casts an unblinking eye over the scene -- in this sense, it is a dramatic improvement over human performance.

Myth Two: Video analytics can improve bad CCTV infrastructure. Most CCTV systems aren't designed with video analytics in mind -- in fact, many are designed without human operators in mind. Cameras are often low quality; views are obstructed by natural or man-made obstacles such as trees and buildings; and scenes are often poorly illuminated at night. Analytics is no "magic bullet" that can see through walls or turn night into day. If the camera doesn't have a clear unobstructed view of the area of interest, analytics will struggle to add value.