More and more people are choosing to install security cameras, with the number going up by 12% every year. Having a CCTV and VMS (Video Management System) installed often gives us a sense of security. The truth of the matter though is that this sense of security might be misplaced or exaggerated. This is because most video surveillance systems still operate at a very basic level; they simply provide real-time raw video footage.
Getting useful insights from the VMS can sometimes feel like accessing the internet without having a search engine. You have to comb through hours and hours of footage just to find what you’re looking for.
Even worse, while the VMS does work in case of major events like break-ins and gunshots, smaller events like theft, leaving a suspicious package and so on can go completely unnoticed. This is because the VMS is being monitored by operators who are sometimes monitoring several screens at once, making it humanly impossible to catch everything that is amiss. This is why VMS systems are limited to either big events or finding footage after the crime to enable investigation.
How AI can remove existing limitations
An effective AI video surveillance system is almost like having a security guard with near perfect efficiency. The AI system can track every action in every video stream, 24 by 7. Not only that, the AI-based algorithms allow the system to identify different kinds of suspicious activity which can then be reviewed by security personnel and action can be taken.
Another major application of AI would be in reviewing surveillance footage during a forensic investigation. Instead of law enforcement officers reviewing hours and hours of footage to find what they’re looking for, an AI-based VMS would allow them to search for people and events. Thus, the effective application of AI algorithms can exponentially increase the efficiency of current surveillance systems.
Why has it taken so long for AI to be adopted
According to a blog post from Memoori, there are two aspects to implementing AI in a video surveillance system. The first part is adding the analytics engines to the video streams and enabling them. This is fairly straightforward and is a one-time effort.
The second part is customizing and training the AI based on the unique site and circumstances. Not only will a home have different security requirements than a shopping mall but a shopping mall may also have different requirements for the food court and for a luxury store. This makes it very time-consuming for AI to actually be configured and implemented effectively. There is a constant need to adjust detection zones and masks, camera angles, perspective settings which causes resistance to the adoption of AI-powered surveillance systems.
However, in recent times that has changed with the advancement in technology. There have been major advancements in semiconductor architecture which are now allowing AI equipped surveillance systems to analyze data much faster than before.
The future of AI in surveillance
The rapid advancements in technology have made everyone optimistic about the role of AI in powering a transformation in video surveillance. This has led to major investments by VC firms in AI-powered surveillance systems.
The release of IC Realtime’s flagship AI-powered surveillance system Ella is another step in the right direction. Ella uses AI to realize what is happening in video feeds and makes everything searchable. With all these positive steps, AI is finally poised to exponentially increase the efficacy of surveillance systems.