Video analytics - the new magic solution?

Jeremy Kimber, Honeywell's EMEA Marketing Leader, analyses the rising phenomenon of video analytics in the security industry.

Like the pharmaceutical industry, CCTV is always unveiling a new wonder cure.  Given the rapid pace of technology development, it is inevitable that new

Rapid growth predicted for analytic software market within the next two to three years

solutions are constantly being brought to market.  However, in many cases these new solutions are no better than previous, proven approaches and can even be solutions to problems that do not really exist.

Video analytics is the latest application that is generating huge amounts of interest in the industry.  Respected industry analysts such as IMS are predicting that the market for analytics software alone will be worth hundreds of millions of dollars within the next two to three years.

There is a strong rationale to support the need for such a new solution.  You don't need to be a part of the security industry to be aware of the huge number of surveillance cameras now in use in the UK - it is a fact constantly discussed in the media.  At present, there are over 4.2 million cameras - one for every 14 people in the population - and the number is still growing.  Generally there is still strong acceptance of the value of CCTV cameras in Britain.  However, given the huge numbers, there is inevitably increasing concern from the public over privacy and the potential risks of intrusion into people's private lives.

Currently, the main benefits of CCTV cameras are seen to be an element of deterrence (due to the perception that anyone engaged in illegal activity is being watched so may think twice), and a proven capability to support the identification and prosecution of criminals after the event.  The rapid responses to the 21/7 terrorist plots and the more recent abortive airport and nightclub incidents are good examples of this latter point.  However, there is a growing desire to see if CCTV can be turned into a more proactive tool.  This is typified by the public address systems linked to cameras in Middlesbrough, for example, which allow operators to reprimand individuals via loudspeaker, helping reduce minor misdemeanours and prevent bigger problems.

Video analytics technology:  advantages for end-users

Video analytics software offers effective way to manage vast amounts of data

Video analytics potentially helps address these concerns and is one reason why the technology is seen as such an exciting new development.  Users of video are looking to find more effective ways to manage the vast mass of undifferentiated data that they are currently receiving from cameras and turn it into useful information.  They are also looking to find ways to turn their cameras into predictive tools that will allow them to spot problems brewing and prevent incidents in a more proactive manner, rather than just filming the events for later investigative use.  By supporting these changes in the use of CCTV, analytics will help address the public concerns over invasion of privacy.  As video becomes increasingly used for targeting real criminal incidents, then in turn, the public should become more willing to accept further increases in the use and number of cameras without a major backlash on privacy grounds.

In broad terms, video analytics uses a variety of rules, which can be specifically tailored both to the scene and the objects being observed, in order to intelligently identify potentially suspicious behaviours.  By using multiple tailored rules and algorithms that will screen out adverse weather, the effects of changing lighting conditions, and non-critical movement activity, analytics systems let security staff focus on real incidents rather than getting bogged down with hundreds of false alarms.  One misconception about analytics is that it is just advanced motion detection.  The intelligence in analytics systems in practice means they are able to offer massive reductions in false alarm generations compared to standard motion detection.

The intelligence in analytics systems in practice means they are able to offer massive reductions in false alarm generations

Humans have an amazing capacity for decision making but are notoriously poor at maintaining concentration levels. A variety of studies have shown that after 20 minutes of watching, up to 90% of the information being shown on monitors will be missed as observers lose concentration.  In analytics systems, the application does the mind-numbing job of monitoring, using the rules and algorithms to screen out unwarranted alarms.  Only suspicious behaviours then trigger the alarms, allowing the security staff to focus on using their decision-making capabilities to identify if it really is a threat and warrants further action or is still a false alarm - potentially leading to a retuning of the analytics rules to continuously improve the effectiveness of the system.

There are a number of video analytics packages available in the market.  Typical applications include:

Video analytic perimeter protection systems

Perimeter protection systems provide back-up to fences, external pirs, seismic systems etc., allowing the user to identify specific areas where intruders will be identified.  Potentially this includes virtual ‘fence' lines that will trigger when an intruder climbs over it (rather than when a guard patrols along it), tripwires that trigger when crossed in specific directions and alert areas, such as nearby roads, which will trigger if a car sized object dwells in them for too long - i.e. if a car stops on the road or pulls over near the perimeter being monitored.  Combining these rules ensures only suspicious behaviours trigger the alarms and not the local rabbit population.

Dwell time video analytics

Monitored area of interest applications allow parking lots, one way streets, doorways and other specific areas to be monitored to avoid cars being left in no parking bays, to identify vehicles or people moving in the wrong direction (up one way roads, up exit only gangways at airports) and to highlight excessive loitering (via dwell time analytics).

Identification of left luggage through video analytics

Left baggage systems allow the identification of objects that h

Use of analytic software within perimeter protection systems

ave been left behind or left stationary for too long, particularly for transport locations although this may not be fully effective in very crowded environments.  Some analytics systems, however, may even identify when the scene is getting too crowded and that they are no longer able to function effectively and need to transfer monitoring back to the security team.

Use of video analytics for market analysis

A lot of the analytics information may also be valuable for marketing teams in companies as well as the security teams.  Many analytics solutions have marketing packages that provide people counting, car counting and dwell time functions (to identify if customers stop by key displays).  Whilst potentially very useful for the client, a bigger question may be whether you as security installers have the right contacts and capabilities to reach and effectively promote these solutions to end-user marketing teams.

Application of video analytics in forensic science

The power of the new analytics platforms also offers a significant improvement in forensic analysis capability.  Recorded video can be fed through the systems, post incident, using the tailored rules to allow faster and more effective identification of participants or events involved in or related to the incident.

In summary, analytics offers the ability for security teams to become both more effective by proactively addressing suspicious activity prior to incidents occurring, and more efficient through improved monitoring performance and speeding up post event forensic analysis.

Evaluation of architectures supporting video analytics

Analytics is going to be an increasing part of video solutions over the next three years and will be as ubiquitous as video motion detection within five years

As well as the range of analytics applications, there are also a number of different architectures that support analytics.  There is a lot of debate about the theoretical advantages and disadvantages of analytics at the centre (on servers or DVRs at the on-site control room or central monitoring station), at the ‘edge' (built into cameras or streamers on site) or with hybrid systems.

In practice, the different architectures will be relevant for different clients with different needs.  End user companies with existing infrastructures who want to upgrade by adding analytics onto some of their existing cameras may find that integrating analytics systems in their control rooms is the most cost effective and flexible solution.  Customers with limited bandwidth on their networks may want to put IP cameras with built-in analytics on site so that only information on suspicious incidents is being sent through the network.  Since the data will have been converted into ‘metadata' by the in-camera analytics, it will also be even less of a drain on the network than normal video.  As an installer, your best approach is to keep an open mind and look for manufacturers who can offer a range of analytics architectures allowing you to offer the best solutions for the individual requirements of your different clients.

Analytics is going to be an increasing part of video solutions over the next three years and will be as ubiquitous as video motion detection within five years. 

 

Jeremy Kimber 
EMEA Marketing Leader
Honeywell