Video surveillance has been a part of the security arsenals of businesses and organizations for decades now. In most public places, cameras are carefully positioned so that security personnel can monitor suspicious or threatening activity from a control room. While it is valuable to have eyes all over a facility with live feeds displayed on scores of screens, this alone does not ensure the security of a facility. It does not ensure that every questionable occurrence will be detected and addressed. In fact, studies have shown that most human beings have a concentrated attention span of only about 20 minutes when engaged in mundane tasks. This time span is even shorter when individuals are engaged in multitasking, such as monitoring multiple screens. Thus, there is immense opportunity for human error when security officers are tasked with monitoring surveillance screens for hours on end. This predicament is well known to the security industry. In recent years they have begun using a cutting edge, borderline futuristic technology to address this problem: artificial intelligence.
Now, when you read the words “artificial intelligence”, your head probably filled with images of a war between robots and human kind. Don’t worry; the AI that we’re talking about is much less complex and much more practical — and safe. Think about the last time you searched something on Google. Let’s say you searched for something random like “laundry soap.” If you were to log onto Amazon.com after that search, there is a good chance you would be presented with a list of recommended items that includes fabric softener, stain remover and dryer sheets. That is practical artificial intelligence. If you have a history of watching TV crime dramas on Netflix, Netflix will likely identify this trend and begin recommending new crime dramas to you. That is practical artificial intelligence. Based on interaction with an operator, computers can identify trends or routines, and adjust to them.
This technology has obvious application to remote video monitoring. Developers have used the function of identifying normality to detect abnormalities. Here’s how it works: surveillance cameras equipped with video analytics software detect the presence of humans or vehicles. Then, the analytics software system used to monitor the area learns, through operator feedback, how to recognize which events are of interest to security personnel. This learning takes place progressively, thus making the system “teachable.” When the system identifies suspicious or threatening behavior in an individual or vehicle, it will alert security to address the situation.
The application of artificial intelligence to the video analytics world has created some wonderful and pragmatic solutions for businesses and organizations that use the technology. Consider a few of the possibilities created by this technology.
For retail stores that struggle with shoplifting, artificial intelligence technology in video analytics has given them some new tools to be able to detect and prevent theft. Many analytics systems are now equipped with face recognition technology, which uses security recordings to map the facial features of shoplifters and other people of interest. When these people return to the store, video analytics identifies their faces and alerts the store manager or security personnel so that they can deal with the situation immediately.
Artificial intelligence can also identify suspicious body language and questionable activity. The criteria for these labels can be established and clarified through interaction between the software and an operator. This feature can be used essentially to predict theft or other crimes before they even happen. When an individual is identified by the system as participating in suspicious behavior, security personnel are alerted so that they can monitor the individual and respond immediately to any security threat.
A common feature of many video analytics systems is a virtual trip wire. This is basically an invisible line determined by the analytics software, which will trigger an alarm when an individual or vehicle crosses it. It is used primarily to ward off trespassers. Since the onset of video analytics, programs have struggled to differentiate a human being crossing the line, which would be significant to security representatives, from an animal crossing the line, which may not be of interest. This dilemma has been recently dealt with through the use of AI video analytics, which have now learned to detect the differences in appearance and body language between animals and humans. It is small, incremental advancements like this that make artificial intelligence an increasingly valuable asset to video analytics.
~ Rick Delgado