You are here

AI in Safety and Security

AI in Safety and Security: Where it has been and where it is going

It’s also a time when businesses are becoming more aware of the need to bring technology closer to their customers.

AI is about making computers imitate human behavior to achieve that goal. Machine learning is an approach to achieve AI, by giving computers the capability to learn. Going further is deep learning, which creates complete machine learning via neural networks.

A brief introduction to AI

AI and machine learning can be traced back to the 1950s. However, it should go more than 40 years before AI and machine learning technology started to flourish. In the last 25 years we have experienced an explosive adoption of AI technology that is driven by big technology steps in ML algorithms and computing power.

AI Timeline Blog

There are three types of machine learning:

  1. Supervised learning that includes a dataset with answers
  2. Non-supervised learning that includes a dataset without answers and learning from observing
  3. Reinforced learning that includes a goal and a reward and learning from actions

Machine_Learning_Transparent

When most people think about AI today, they picture it as supervised learning. For example, an individual could feed a computer with 2,000 pictures of different animals. The machine will learn how cats and dogs can do, based on the photos, for example, and then be able to classify the pictures based on those labels. That is supervised learning.

Non-supervised learning is the process of just feeding data to a machine and the machine tries to categorize and group it and learn from observing it. This is the way that human babies learn.

How is AI technology being used in the safety and security industry today?

Every security organization can benefit from AI technology, which provides faster computing and insights, better data security, and efficient control over continuous operations. It has many applications in the security industry, and many global businesses are already reaping the benefits of AI.

Specifically, with video, AI can generate real-time alerts when an object is identified and tracked through a pre-defined area. Operators based in a control room can get an alert when there is a potential security issue.

In an airport, AI can help with wrong-way detection with a restricted area. This responsibility is usually given to security guards; however, AI technology can reduce the number of security guards that are needed, which reduces costs while also increasing the accuracy of a security system.

In a retail store, AI can identify a left behind or an abandoned object and alert a security team to act.

Where do you see AI technology developing in the next few years?

For the security industry I would like to highlight three areas where I see AI technology developing and being used more in the next few years. These are:

  • Audio analytics
  • Digital policing
  • Human – machine collaboration

Unlike vision, audio waves do not require direct line of sight to be detected, as audio bends corners. This opens new use cases when it comes to detecting critical situations. Take the example of a person breaking into a car in a car parking garage. He would most likely do this in an area that does not have security cameras. However, an audio analytics solution would still recognize the event and alarm the security guard to take action.

Digital policing provides new capabilities to both police unwanted behavior as well as tracking down criminals. Take the example where a security camera has taken a photo of a bank robber and the police use facial recognition together with the database of driver license photos as a search engine to identify the criminal.

We already see digital policing being used in China and in a lesser extent, in the US. However, criticism has been raised around digital policing, and some say digital policing could lead to a “Big Brother” society where everyone’s small mistake or behavior is policed.

In addition, there have been several incidents where AI technology has proven to take wrong and biased decisions (For more reading see sources below).

Still, the analytic skill demonstrated by the most advanced AI technology is dumber than a rat. That being said, the machines and AI have a set of unique skills such as never being tired, the ability to observe millions of inputs in parallel, and a massive memory. This makes the machines and AI a super partner and assistant when it comes to improving our efficiency. Human machine collaboration will give more optimal processes and efficiencies where we can benefit from the analytic skills of the human brain together with AI. We have so far only seen the dawn on how this human/machine interaction can be with Siri and Alexa, and the Tesla autopilot.

In my next blog, I will discuss how AI can help companies tap new sources of data for analytics.


About the Zenitel CTOZenitel Thomas Haegh 1024

Thomas Hægh joined Zenitel in 2003 and became CTO in 2009. He received his masters degree in electrical engineering in 1994 from the Norwegian Institute of Technology.

Thomas has 25+ years experience in hardware and software design developing IP communication solutions.

10 Questions to Ask Before Selecting an IP Intercom Solution

How Audio Adds ROI to Security Programs

Zenitel, CHeKT Announce Product Integration

Zenitel, Feenics Announce Product Integration