Video content analysis (VCA) can save CCTV operator time and, for instance, spot suspects quicker. But do end users expect too much of VCA if they want to call up every sighting of a man with a red jacket?! The expectations and realities of VCA were aired at a conference. Roy Cooper and Mark Rowe attended.
Projects with VCA need to show a return on investment (RoI). Installers and manufacturers have to manage the customer’s expectations, and deal with any problem of high false alarm rates. There has to be ease of installation. These were the issues highlighted at the IMS Research conference a year before, Alastair Hayfield of the market research firm said. What has changed? There’s been some progress in educating the security industry about VCA, he said: “We think in general installers and integrators are better informed about what the technology can do, and the advantages and disadvantages.” The VCA sector has shown, he went on, that real-world problems can be solved by such software, giving a genuine RoI. As for managing expectations, he admitted the early years of the product had seen hype; but there was progress there, too.
However as for vendor stability – something the customer would want to see – VCA companies like any others have found the last year or two difficult; and it’s several years since VCA vendors received European Union funding. And false alarms remain a problem, when systems have not been configured properly. Nor, he suggested, has the last couple of years seen major advances in the technology. So, you might wonder, why a two-day conference in London devoted to VCA, at the same venue (Paddington Hilton) that the conference ran in four years ago? In a word, the potential world market for VCA. While 2009 saw no real development over 2008, and 2010 is forecast to be pretty slow for growth, as the economy picks up the researchers expect to see VCA sector growth return, though for the foreseeable next few years fairly few CCTV devices may feature VCA. IMS has tracked 1000 tenders around Europe and North American in the last six months, and only about one per cent were requesting VCA at an early stage. (Some of the other 99 per cent may go on to use it, as Alastair Hayfield pointed out.)
So why the persistent interest in VCA? It could be that the typical applications so far are large. Alastair Hayfield went through them: first, airports, where new security requirements include scanning baggage. Security spending at airports has trebled over the last six years, for instance on guards at fixed posts. A VCA product could monitor people flow, to detect specific behaviours, using a ‘rule-based software engine’. If a behaviour set as suspicious by the user is detected – such as someone going into a prohibited corridor, or entering an ‘exit’ door – an alert can go to a PDA. It’s not, as Alastair Hayfield said, an ‘out of the box’ installation, and complicating the fitting is that an airport is a 24-7 site and you might only be able to test the product with low volumes of people flow, and extrapolate what would happen with peak times of queues. Success can, and has, meant less spending on guards; for instance, an airport can spot queues building and send more guards to start extra lanes. That’s not only good customer service but prevents fines, if service level agreements set penalties if passengers have to wait for longer than 15 minutes, for example in a queue for a security check.
Metro case study
Another case study that Alastair Hayfield mentioned: a metro faced with graffiti and vandalism in tunnels. The metro wanted to use its analogue CCTV cameras with encoders already there. Rack mounted computers at each stations were able to handle eight cameras, images transmitted to a central server. Before the VCA went live, it was tested with trains running and not, various lighting, and litter blowing through station platforms – all the man-made and natural things that can cause VCA false alarms. The configuring in other words took time. The installers only had specific bandwidth to work with over the network, even if several alarms went off at several stations.
Out of hours
Alastair Hayfield gave a school and university case studies also. Because an unnamed school’s grounds were open out of hours to the public, it suffered serious and costly vandalism. So how to discriminate between people with a reason to be on the school grounds, and people doing damage? The VCA provided loitering detection and tight alert zones, so the school saw a sharp fall in damage, and a cost saving on repairs. Hence the system paid for itself in a year. As for the university, it sought to make best use of its security guards. Again, the VCA product manufacturer and installer had to work on fitting something not ‘out of the box’. For instance the guards were not necessarily computer-savvy and needed training to get to like the product and what it could do for them. In conclusion, Hayfield advised installers that they should benchmark VCA against what products (such as infra-red lighting with their own false alarm rate) are already in place, as a way to prove that VCA will do better. That is one way to manage the expectations of the customer. You must also have buy-in from the people using the product (who are not necessarily the same people as the client).
Who was there?
Exhibitors included ASL Vision, Bosch, and Ground Source Video, while in the audience were Dominic Bruning of network camera manufacturer Axis Communications, and Peter Strom, chief of March Networks. And people from Siemens, Saab, NICE Systems and Alcatel-Lucent; and from VCA Technology, Keith Waterhouse and Geoff Thiel. To sum up, as the conference heard, few projects have large numbers of CCTV cameras, with analytics. Nor has the technology become ordinary and everyday enough; the sector is still searching for the ‘killer app’. Yet its appeal remains; and not only for security uses. As one speaker put it, Soren Svensson, of the Swedish Defence Research Agency (FOI), VCA offers ‘alerts from subtle signs on boring video screens’. In other words, VCA can automatically detect crime or other threats that control room operators may not, when faced by banks of monitors where little happens. Svensson, by background an engineer, went into the Prometheus project, financed by the EU, and the ADABTS project, that began last year and included from the UK the Home Office and BAE Systems. Prometheus stands for prediction and interpretation of human behaviour based on probabilistic structures and heterogenous sensors. Put another way: trying to track multiple people (because outdoors and in people do not behave as smoothly as in laboratories!) and interpret the human action. If someone stationary is viewed on CCTV for example, is that person waiting for someone; or for an opportunity to assault someone? As for non-security applications, VCA and other sensors could save on healthcare staff. For instance, if a patient goes to a bathroom and is there longer than a set time, if sensors do not pick that person up in the corridor or their room, without invading privacy you can deduce that the person may be in difficulty in the bathroom, and call or attend. Other work is going on in analysis of crowd behaviour. That includes defining abnormal behaviour; a loud crowd in a football stadium may be normal cheering, but a threatening mob if at a railway station. As Svensson stressed, there are legal and cultural questions around use of VCA, besides the economic ones of saving on guards or nurses. He speculated whether an air traveller in the 1950s would have accepted a luggage or body search to get on a plane. Now it’s accepted. Might VCA – for example used for people counting – gain the same acceptance?
This article was originally published in Professional Security Magazine. If you would like more information on subscribing to this magazine please Contact us.
If you would like some more information on Video Analytics, please Contact us.