Artificial intelligence (AI) is making rapid inroads into the physical security sector – but the steep learning curve and technological hurdles could pose a challenge to systems integrators (SI).
Artificial intelligence (AI) is making rapid inroads into the physical security sector – but the steep learning curve and technological hurdles could pose a challenge to
systems integrators (SI).
According to a report from Research and Markets, AI will be a key driver of growth in the global physical security market, with the firm projecting a CAGR of 7.3 percent from 2018 to 2023. AI video analytics are also expected to boost growth in the video surveillance market, which Memoori predicts will see a CAGR of 13.43 percent to 2023.
But AI implementation could be slowed if SIs are unable to keep pace with changes in the industry. According to Jeff Whitney, VP of Marketing at
Arecont Vision Costar, AI systems will still be new technology as they find their way into the general security marketplace, and as such their strengths and weaknesses will not be known until SIs have significant experience with them.
“Just as when IP camera technology came on the scene, there will be a real learning curve to get most of the technology and to understand its limits, not just its benefits,” Whitney said. “Hands-on experience and real-world implementations will be a driving factor in forming realistic expectations for the customer from the integrator selling and installing the system.”
Other challenges in the use of AI
Experts across the board have highlighted other challenges to the implementation of AI. Prominent among them is that there is no clear consensus on what the term AI encompasses. We are still at a stage where machines need to learn, and this requires data sets that can train a computer before it can be put to use.
There are also concerns over the computing power that is required to run deep learning algorithms that could increase the infrastructure budgets for customers.
In a recent blog post, Martin Gren, co-founder of
Axis Communications, said “Memory, processing power, and power consumption are the biggest challenges within AI, especially for deep learning applications.
“Today’s systems require vast amounts of space to store data for learning, either in the
cloud or on a server with multiple GPUs.”
He added that human experience still beat AI in some cases. For instance, AI can detect a person running but not why he is doing so. Is he running to catch a bus, or has he just robbed a bank? Knowing such information could be key to identifying security issues.
The quality of images captured will also have an impact on how AI processes data. Cameras should be able to capture images in different situations. If they get visuals wrong, a computer could draw wrong conclusions.
What does AI promise?
Challenges aside, AI does remain the clear road ahead for the security industry. Whitney points out that although video quality continues to improve, bandwidth consumption and video storage is limited in its availability and also remains costly. Even as compression technology reduces the need for both, there is still a large amount of video to store, either locally with a traditional surveillance system, or in the cloud.
“AI offers the promise of making better use of video imagery and being able to extract more data out of a frame than a human operator might,” Whitney said. “Thus, AI will reduce the amount of video that needs to be transported and stored, while providing increased value from that video.
“By being able to tie in other sources of info such as from other sensors or applications like identity management, heat monitoring, people counting, queue counting and access control, the overall system can potentially deliver much more useful information to the business or organization beyond the traditional security requirements. Only an AI-based system could effectively manage all of that information and sources.”
In short, from expertise issues among SIs to technological aspects, the challenges that AI poses are not minor. However, the potential advantages of intelligent machines for physical security are too significant to be ignored. AI is still a work in progress when it comes to physical security but it’s clearly going to have a major impact on the industry.