Artificial Intelligence & Video Surveillance: Measurement Vs. Recognition

Posted on Categories Artificial Intelligence

Stuart Rawling is VP Market Strategy at Pelco, a global leader in the design, development, and manufacturing of predictive video security solutions. Stuart recently sat down with us to discuss a range of topics related to video surveillance and technological advancements thereof. We are excited to be able to share a few of Stuart’s thoughts regarding the industry, artificial intelligence in the security and surveillance sector, and how neural network technology affords organizations newfound opportunities for operational efficiency.

As the Director of Segment Marketing at Pelco, Stuart analyzes the needs of end-users in conjunction with Pelco’s products and services. Essentially, Stuart ensures a strong correlation between the operational inefficiencies our clients experience and the surveillance solutions Pelco implement and support.

Where Is The Industry Today?

In speaking with us, Stuart highlighted the immense change the industry has seen over the past few years, with specific nods to advanced imaging, video storage, and the technological capacity to see disparate systems integrated with one another like never before. Said Stuart, “Moving forward, we are seeing a lot of industry buzz around things like artificial intelligence and deep neural networks, and facial recognition amongst other things. And it is a very exciting time for the industry because we are starting to get some capabilities that we’ve dreamed of…now we are starting to see those things become a reality.

So what does this mean for those of us in the security and surveillance industry? What does it mean for those of us who are tasked with protecting people and assets on a daily basis?

Artificial Intelligence & Surveillance Technology

The most pertinent way to answer these questions is to look at recent technological advancements in artificial intelligence, an area where we’ve seen hundreds of years of mathematical progress applied in computer machine learning. Over roughly the last half-century, progress was slow going — until a decade ago, when the industry began to tap into deep neural net technology, which Stuart describes as essentially “a science of replicating the way our human brains learn in a computer.

Stuart continued, “And what that allows us to do is to take images that we generate in our cameras and systems, and then apply them through this convolutional mathematical model that allows us to teach the computer how to recognize things.

A Qualitative Difference: Measurement Vs. Recognition

In the past, the surveillance industry has had challenges with capturing consistent inputs if there were environmental disruptors. For instance, a security camera at an airport terminal might have struggled to get repetitive alarms during a storm because the objects it observed were only measured, rather than recognized.

…It was looking at the things it saw; it measured them, compared them to the background, compared them to the ambient light and things like that, and then it made the decision about whether it was an object of interest. And the object, it was not really recognized; it was just measured. And the difference is that inference that it does when using measurements is only as good as the way the camera gets set up, the ambient light, and the way all of these things come together.”

With deep neural network technology, we are teaching the computer not to measure things, but to recognize things. That’s the big difference.”

This is nothing short of a paradigm shift for our vertical. Modern video management systems can now examine an image and classify people, vehicles, and even different kinds of behavior in humans. Just like before, this information is grouped together in a VMS to form events, but the key difference is in how we are able to utilize these events in novel ways.

Stuart sees this progress as nothing short of auspicious: “just improving the way we generate data is good, and we get much more accuracy, and much lower levels of false alerts — and that’s all good — but when we start to bring these things together, when we start to look at an entire operational security picture across a facility, and we start to apply that learning technique to the entire operation, that’s when we can start to really start to make a difference in securing operational facilities.

A final example: as an operator, you see a perimeter alarm set off, then movement on a camera of a person of interest in an area of concern, all before being alerted to an unfamiliar license plate number entering the premises earlier that day. Modern surveillance solutions developed by Pelco can automatically collaborate and alert you to this information in real-time.

As someone whose expertise is in identifying the operational needs of end-users in order to link them to relevant solutions, Stuart’s optimism is echoed by all of us at Pelco. At the end of the day, we are here to provide pragmatic, intuitive solutions to help clients make better decisions, improve efficiency, and ultimately be more profitable.

Browse Pelco’s video security solutions for a range of applications at your convenience.