By Gregory Hale
Seeing is believing. Or is it? When it comes to physical security and witnessing what is occurring before your very eyes in a video application, sometimes humans don’t have the real answer or they need help.

That is where artificial intelligence (AI) and machine learning come into play – and that is not something that is coming way in the future, it is becoming more mainstream today than ever before.

“When machine learning has the right data, training and algorithms to support it, it can do its job which is to look for patterns,” said Brian Baker, vice president, Americas at Calipsa, a cloud-based platform provider that uses machine learning to reduce the number of CCTV alarms, during a presentation last week at the GSX+ virtual conference, entitled “AI Is Powering a New Generation in Video Analytics.” “Until the early 2010s, pattern recognition didn’t work so well when it came to images. In ten years, technology has come a long way and it is now much easier to train computers and spot patterns in images in video. However, machine learning still requires a lot of input to make it truly effective.”

Video analytics are becoming more powerful and accurate than ever before. Gone are the days of staring at a video screen for hours on end to see if there is any movement or activity going on in a certain location.

Schneider Bold

Now with artificial intelligence and machine learning, operators can truly get a solid grasp on what is going on at any given moment.

Machine Learning
Simply put, AI is any technique that allows a computer to mimic human intelligence. Some of the techniques include logic, if then rules, decision trees and machine learning. Machine learning is just one type of artificial intelligence, Baker said. Machine learning algorithms use statistics to find patterns in vast amount of data and then apply those patterns to make predictions. The more data algorithms work with the better they are at spotting patterns.

Deep learning is a subset of machine learning where the algorithm trains itself to take in vast amounts of data.

Deep learning depends on neural networks so called because they were inspired by how neurons work in our brains. The human brain can solve a complex problem, but smaller neurons are each responsible for solving specific smaller parts of the problem of the whole. Similarly, a neural network is made up of many layers of computational nodes, each node is responsible for solving a small part of the task at hand a bit like the neuron in the brain. This gives computers an enhanced ability to spot patterns as all the nodes work together to process much larger amounts of complex information to make a prediction.

“When humans watch video, we take for granted the highly sophisticated process taking place between our eyes and our brain in fractions of a second without any effort from us,” Baker said. “Something as simple as watching someone sit in a chair involves multiple levels of understanding – what a person looks like, what a chair is, and the downward motion we are watching is the transition from standing to sitting.

Machine learning takes in the information it gleans from video and is able to extract meaningful information from it.

“We use a machine to analyze video and turn that video into data,” Baker said. “It can classify and detect objects of interest and understand the movement and behavior of the objects. In most cases an alert or an alarm will be generated based on the identification of the person or vehicle based on its behavior.

False Alarms
The problem is those alarms are usually false alarms or nuisance alarms.

“Until now, guards and control room operators have had to monitor continuous video streaming or view the alarms security cameras send through,” Baker said. “In both cases, they could be using their time and expertise much more efficiently. Guards and operators are on the lookout for genuine alarms; this is video that contains activity of where people are not supposed to be. But we know 95percent of alarms are nuisance alarms. What if you can train a computer to filter out all the false alarms?”

That is where AI and machine learning come in.

To show a real-life application, Baker presented a video where one company that was using AI and machine learning to reduce or eliminate nuisance alarms.

“We want to reduce the amount of false alarms which should free up more manpower so they can work with customers,” said Aleksei Bulavko, research, development and production manager at Vancouver Fire and Radius Security. “There has been a long wait for machine learning analytics on the edge and where you can transfer the data to the cloud. It is there, it is just a matter of using it right. This security is not a playground, we don’t have room for any mistake, we are dealing with people and company assets and sometimes people’s lives.”

Believe it, advancing security through AI and machine learning is here to stay.

“Machine learning enables us to do things better and faster than we could have done before, and over time it is only going to improve,” Baker said.

Pin It on Pinterest

Share This