Analytics for the real world

The over-ambitious claims and promises that may have shrouded analytics some years back have been replaced by a more sober and solutions-based approach to the technology, as SME discovers.

Video Content Analysis (VCA) – otherwise know as video analytics – is certainly an area that generates considerable debate. The wider deployment of VCA, which supports the automatic analysis of CCTV images has, some would argue, been held back as a direct consequence of overly ambitious claims in the early days that were simply not matched by system performance. Thankfully, technology has moved on considerably in recent times, whether it be at the edge or located centrally, and many of the initial operational issues are starting to be addressed. There is also, crucially, a more realistic appreciation of where video analytics and its associated algorithms, can be deployed to best effect.

Market potential

A new report published by IMS Research entitled: ‘The World Market for Video Content Analysis in Security and Business Applications’, sheds some light on what the future holds, forecasting that the market for chargeable VCA applications will continue to grow quickly with sales related to video surveillance devices reaching $600 million by 2015. The sectors where take-up is expected to be strongest include: retail, energy and utilities and transportation. For the Middle East and Africa, IMS Research estimates that the market for PC-based VCA software amounted to $3 million in 2011 and for intelligent video surveillance devices the figure was $5 million.

The report’s author, Jon Cropley stresses that a clear divide is now emerging between free, lower-end, applications, like motion detection and camera tamper, which have been embedded in video surveillance devices for sometime and higher-end chargeable solutions.

Explains Cropley: "Detecting an intruder inside a building from a few metres away requires a much less advanced solution than detecting an intruder in an outdoor environment from a greater distance and in conditions where there can be significant movement from peripheral features like branches moving in the wind.”

Cropley believes that it therefore makes sense for a charge to be applied for higher-end VCA to recover the high development costs: "The reality is that it is unlikely that suppliers will get ROI on this outlay in an increased price for every device they sell. In addition, many VCA applications are quite niche so without targeted charging the majority of customers would be paying for a feature they would not use.”

A cautionary tale

For his part, Ivo Drent regional director at IndigoVision, signals a cautionary note with regards to video analytics. In terms of a technology adoption curve he believes that VCA is still very much at the bottom of the ‘chasm.’ "A couple of years ago we were seeing a huge amount of hype about video analytics with new companies appearing with tens of millions of dollars behind them but many have fallen by the wayside. It simply hasn’t been the goldmine that some may have initially thought. The reality is that we can make the basic stuff work pretty well – like tripwire – but the promise that you can effectively detect a person acting suspiciously, such as when entering a subway, is still just fantasy. People don’t behave as straightforwardly as that.

"At a practical level before opting for a solution, potential users need to consider what they want to achieve, that it is really doable, and account for elements like camera positioning,” explains Drent.

"Where a camera is actually sited can make the work of the analytics algorithms 10 times easier (or harder). Taking counterflow detection in an airport environment as an example, when a camera is looking down at the scene it is really straightforward, whereas with a camera viewing the scene from the side, or at an angle, there is the issue of perspective and objects that are in front of each other come in to play – this is when things start to become really tough.”

IndigoVision as a company deliver systems that they have their own analytics and also integrate with third parties. "We supply basic analytics – motion detection analysis – as standard on our products, people don’t have to pay extra for this. I would guess that maybe twothirds of customers use it. In terms of advanced analytics the take-up is still pretty low, probably about a five per cent. The contrast between the two figures really tells you where the market is at the moment.”

The sort of image processing involved in video analytics is, by its nature, very computing intensive. Drent’s view is that the only way to do it properly when looking at what is happening live is at the edge of the network in the camera or right next to the camera. "If you are bringing everything back to a central point and then analysing you have bottlenecks and it is not scalable. There is the fact that if the central server goes down the whole thing goes down.”

Where Drent does envision a role for centralising analysis is in the post processing of evidence, an area where IndigoVision has taken a lead with a specialised forensics module. "If you have huge numbers of recordings and want to do some data mining to pull out all of the relevant incidents that needs to be done centrally.”

Tailored approach

Roger Decker, director of solutions and marketing at Siqura is a strong advocate of application specific VCA. "At Siqura we are focused on providing the best possible detection algorithms for very demanding applications. This shift from generic solutions to more tailored options is where our efforts are concentrated in areas such as traffic management and the protection of critical infrastructure.”

One project that Siqura has been involved in that showcases its capabilities is an Intelligent Transportation System (ITS) for Dubai Roads and Transport Authority (RTA). Here sophisticated detection algorithms are being applied to video to detect an array of events from stopped vehicles to drivers going the wrong way.

In terms of where video analytics should sit in the surveillance infrastructure, Decker explains that Siqura also advocates the distributed route. "We put the extra hardware in the camera or an encoder. Having analytics at the edge means that this is where the number crunching can be done and the metadata sent back to a central location, which presents a fraction of the amount of data compared to streaming the complete video.”

Siqura’s commitment to video analytics on the edge is reflected in the recent launch of a camera platform using a dedicated DSP (Digital Signal Processor) that enables analytics algorithms to be embedded.

Traditional users

Peter Biltsted, sales director for the Middle East at Milstone reports a marked increase in requests for VCA. "Most of the demand we are seeing is coming from the public sector and ports and airports – the traditional users of video analytics.”

In terms of specific projects in the Middle East, Biltsted says that Milestone has provided one of the biggest applications of video analytics in the region. "Ferrari World in Abu Dhabi is on an impressive scale and has more than 700 channels of analytics. At the theme park the analytics employed ranges from people counting to tripwires and zone violations. Basically this is working with the security system, ensuring, for example, that crowds are not going in the wrong direction or moving into areas where they are not meant to be.”

Biltsted believes that there are still some factors holding back the wider roll out of video analytics. "A main commercial issue is the price point for analytics when weighed up against the relatively low cost of security guards. Another constraining factor is that the industry needs to be much better at getting the message across regarding the return on investment provided by video analytics. In addition, the cause of video analytics has not been helped when suppliers have provided cumbersome solutions, requiring considerable calibration, where the ease of installation and usability has not been there.”

Milestone is a long-time supporter of an open platform approach to video analytics and as a consequence offers a framework which is capable of integrating video analytics tools from different manufacturers. "Customers can pick the video analytics that they want and are not forced to work with a certain analytics provider,” explains Biltsted.

Considering the future, Biltsted expects video analytics to expand in 2012 and 2013 and feels that the companies that will ultimately be successful are those who actually have local support: "Without the right resources in place on the ground vendors might make a sale here or there but they are not going to be able to sustain a presence in the marketplace.”