Video Analytics: The People Have Spoken
Security industry is increasingly comfortable with analytics, cloud.
ONE of the key things to come out of Security 2019 Exhibition in Sydney last month was implied by the winner of ASIAL’s People’s Choice Award – NiroVision’s cloud-based Mobile First face recognition solution.
There were plenty of other capable face recognition solutions at the show, and loads of analytics as well, but it was the growing comfort installers, integrators and end users clearly feel with cloud and with video content analysis based on deep learning that was so interesting.
Speaking recently with large end users and enterprise-level integrators, a constant refrain has been an active interest in analytics, with face and object recognition capabilities being particularly favoured. There are a number of reasons end users want analytics – efficiency is a big one. The ability to find faces, as well as clothing, licence plates and all the rest, across a network of hundreds, even thousands of cameras, is so compelling it’s impossible to ignore.
The efficiencies aren’t just about opportunity costs won through saved time. For serious security applications, being able to recognise and track threats or vulnerabilities – such as missing children – at extreme speed is the key to successful operations. Something else that’s at the heart of analytics is the importance of maintaining the operation of major infrastructure – that might be a railway station, an airport, a freeway, or a city street. And this is a priority not only for infrastructure owners, but for law enforcement bodies tasked with protection and coal face management of cities. Faced with emergencies, they want to know.
Efficiency is about money, too. End users in the private and public sector have millions of dollars tied up in their surveillance systems. The latest solutions are extraordinarily capable and the data they generate is solid gold. But to deliver on their promise, this gold must be mined quickly and accurately to deliver the situational awareness that leads to an expedited response.
Wandering around the show last month looking at VCA solutions I couldn’t help thinking about the pioneers of analytics – companies like Briefcam and Avigilon, whose solutions have slowly and surely shifted the expectations of end users. Today, security managers and operations teams don’t want to drown in video streams but be informed and empowered by them – and nothing delivers this capability in the way analytics delivers it.
“The entire notion of siloed security systems is being unstitched at the operations end, where managers want more, faster.”
Something else we’ve noticed looking at case studies and product reviews this year is that CCTV systems are being federated and integrated on a scale we’ve never seen before. The entire notion of siloed security systems is being unstitched at the operations end, where managers want much more, much faster. And there’s a lateral element to this hunger driven by the possibilities of technology, as well as by the growing expertise of integrators who have the capability of wrangling huge interconnected networks. We are starting to see overlapping ‘super systems’ of cameras, offering granular access and massive potential, but this power comes at a price.
Experienced control room managers know that there’s a point at which adding cameras to a video wall no longer enhances situational awareness but creates noise, slowing decisions and attenuating response times by making the system much higher touch. It’s this confluence of interconnected CCTV networks and the operational constraints of monitoring them that’s going to drive the uptake and functionality of video analytics to the next level, in my opinion.
Deep learning is playing a major part here – it’s delivering reliable functionality with considerably lower processing demands and doing so with none of the old-time programming nightmares developers used to face. With deep learning, systems learn their environments from the moment of power up and do so at astonishing speeds.
Just how video analytics like face recognition will play out in the user market is interesting – the general public in many countries is discomforted by face recognition in public surveillance applications and this may lead to tweaks like face redaction in some applications. But it’s impossible to imagine a future in which analytics doesn’t have a huge impact on security management solutions of all kinds.
As it stands, most large CCTV installations are not running anywhere near full potential in terms of data gathering and report generation and that’s because operators simply cannot drive every camera down to the pixel level. But video analytics can – and end users know it.