If one was to identify the one major trend in security over the past 12 months, it would be advanced analytics engines that are often referred to as artificial intelligence or AI. This article examines how these analytics can help improve security, and why some players take issue with the term “AI.”
Compared to the often overhyped and overpromised analytics from the earlier days of security, today’s analytics engine have reached a certain level of maturity and have proven effective in objection recognition, smart search and other applications. This has been possible thanks to several factors, including higher computing power, more advanced algorithms, wider availability of data that systems can be trained with, and how much you understand the demand of a vertical market.
The end results are solutions that help users achieve further efficiency and situational awareness. The adoption of artificial intelligence (AI) along with video analytics technology and software is significantly driving the market growth and is contributing substantially toward keeping the market growth prospects upbeat. The use of this technology is on the rise as it has facilitated a more effective and convenient access to images and videos, thus increasing the overall efficiency in making decisions and detecting threats. The integration of AI in security systems is instrumental in providing smart perimeter security solutions with complete automation to detect intruders and take actions on the basis of instructions fed into the system upon analyzing the situation at the incident site.
AI is definitely taking inroads stronger and stronger all over the security market, simply because installations in general are getting bigger and bigger, with more sensors and cameras, and thereby it’s impossible for the human to keep track of all the information coming in. Therefore, in order to get maximum outcome out of your security installation, you have to use AI in order to capture or interpret the data from your security installation in a good way.
The fact is today there are more cameras and recorded video than ever before, which means security operators are faced with the challenge of keeping pace. On top of that, people have short attention spans. However, AI is a technology that can help overcome this challenge as it doesn’t get bored and can analyze more video data than humans ever possibly could. The role of AI in security is transformative. AI-powered video management software is helping to make security operators more efficient and effective at their jobs. By removing the need to constantly watch video screens and automating the ‘detection’ function of surveillance, AI technology allows operators to focus on what they do best: verifying and acting on critical events. This not only expedites forensic investigations but enables real-time event response as well.
One example where AI can be quite useful is smart search, where instead of looking at hours upon hours of recorded footage, the user can enter search queries and get the relevant video quickly.
And these analytics go way beyond smart search. AI-enabled facial recognition, for example, can be quite effective and accurate in matching a person against the database even if that person is wearing a mask or disguised.
The promise of AI/deep learning is significant improvements in performance of existing applications, along with introduction of advanced new features that were simply not viable earlier. Whether the challenge is video detection of a disguised face in a crowd, or extracting a telltale signature of a security breach from high ambient environmental noise, AI comes with the promise of higher performance and an exciting future beyond.
Taking Issue with the Term ‘AI’
Some vendors and consultants we spoke with took issue with the term “AI.” According to them, the technologies being used today are at best advanced analytics, not artificial intelligence which they said was still far away. People must be cautious about the word AI, because there is a lot of hype claiming that AI brings huge advancements. True AI is very far off, and will not give us the flying car, or even the self-driving car just yet.
These are ‘advanced analytics,’ not ‘artificial intelligence. AI is much broader term and includes capabilities far beyond recognizing objects and classifying them. AI means that computer literally ‘thinks.’ In order to make it think you need to combine many complex technologies such as machine learning, reinforcement learning, generative adversarial networks and augmented data learning.
According to them, today’s technology lacks a predictive element – for example determining someone is about to do something bad because he/she is exhibiting behavior uncharacteristic of his/her profile.
There’re plenty ‘self-learning’ algorithms out there. And they came to life even long time before current promising algorithms based on neural networks. These self-learning things that are capable to capture objects and compare their speed with average speed of objects in this scene are really primitives. You hardly can expect from them that they will automatically recognize suspicious behavior of a person. To the moment, neural networks can recognize exceptionally well any static objects you can imagine: from faces to objects on the x-ray scanner. There’re also 3D convolutional neural network that have just appeared recently. These algorithms allow evaluating movement’s complex pattern in order to classify objects behavior. But there’s plenty of work to do and issues to solve before this technology can be fully applied in commercially available security systems. As soon as it happens we’ll be able to start making preventive surveillance system.
AI has demonstrated increased capabilities to solve ‘traditional’ problems, such as facial recognition or person/vehicle detection. AI-based products have not yet been widely shown they can make more complex determinations, such as ‘is this person acting in a suspicious manner’ or ‘is this activity normal for this person, this scene, this environment.’
However, this is not to say that these advanced analytics – regardless of how it’s called – are intrinsically of no value. I would notice that advanced analytics alone is also really complex and powerful technology. It opens up tremendous perspectives for almost every aspect of life. Surveillance systems are just tiny part of it.
Machine learning and statistical reasoning in programs which give the illusions of intelligence are providing usability and functionality improvements to our end users. Recent advancements in machine learning techniques, in particular, in deep neural networks, have become available to everyone.
Adapted from: a&s Magazine
