Video analytics is a fast growing technology. Anthony Hildebrand explains how it can be used for security and to gather valuable data.
They might attract their fair share of Big Brother-related complaints but with most retailers using them for security, shoppers are used to seeing CCTV cameras on high streets.
Despite this, a CCTV system is still often seen as a ‘grudge’ purchase by retailers - an added cost that doesn’t necessarily prevent shoplifting, and doesn’t add to the profitability of the business. Plus, having a surveillance system installed means paying someone to monitor it, in addition to any shopfloor security staff.
As with many areas in retail, though, things are changing.
The advent of video analytics - the ability to analyse the footage recorded by in-store security cameras to produce useful data and customer insights - means that retail surveillance is now more than just a tool for loss prevention. Marketing, customer service and IT roles can all benefit from the information gleaned, but only if the technology is used effectively.
How technology has changed
Traditional CCTV systems were analogue, which means that images are transmitted via pre-digital TV-like cable, and recorded to tape, or digitally encoded and recorded to a digital video recorder.
But the trend in recent times has been towards IP-based systems, which sit on an organisation’s computer network, whether pre-existing or dedicated to this function. IP stands for Internet Protocol, and refers to the addressing system that locates computers and other devices on a network.
A traditional analogue system would require the installation of three sets of cables, for power, video and camera control. One of the advantages of a networked IP system is that a single Ethernet cable can provide all three of these, reducing installation costs for labour and materials, as well as ongoing maintenance costs.
An important thing to remember is that IP security is not ‘internet security’. That is, it’s not about opening up security devices to the web - it’s using an IP address-based network, or local area network, which is similar to using an internal intranet. This new technology has been growing in popularity - industry analyst IMS Research predicts that 2013 will be the year when network video surveillance equipment sales overtake analogue video surveillance equipment sales.
IP-based video makes it easy to send images and data between sites, which is ideal for retailers with a number of different stores. Similarly, video content can be delivered to mobile devices and laptops easily, allowing sites to be monitored and alarms responded to around the clock.
IP video feeds can be encrypted to enhance security, and users can be allocated different levels of access, offering more security for less money. And like a PC or mobile phone, online software updates allow cameras and other system elements to be upgraded and maintained remotely, which extends their useful lifetime.
Some of the major players in the manufacture of IP CCTV cameras are familiar names, such as Bosch, JVC, Panasonic and Sony, but there are also a number of prominent specialist network surveillance firms, such as Avigilon, Axis Communications, Dedicated Micros, Pelco by Schneider, and HD specialist Arecont Vision.
Picture quality is important for analytics programs to work effectively, so HD cameras are fairly common in this sort of deployment. The flipside is that HD images require significant amounts of storage space. Analogue, then, still remains an option. Samsung is one of the biggest CCTV manufacturers in Europe but was relatively late to the IP camera game, and thus is stronger in the analogue market.
An interesting recent development is that of the panoramic, 360-degree camera. It sits overhead and takes a very wide-angled view, which is then ‘de-warped’ so it looks familiar to the human eye. Manufacturers, such as the UK’s AMG-Panogenics, which produces the PanoCam360, say one of these cameras can take the place of a number of conventional cameras. They have also helped develop the role of video analytics.
What is video analytics?
Essentially, video analytics is software that automatically analyses moving images to detect and determine whether something the user wants to look for is happening. Users and software manufacturers generally determine what the analytics program is looking for, though this isn’t always the case.
Analytics software can reside on the cameras themselves - known as residing on ‘the edge’ of a system - but this only works for less complex functions, such as people counting. More processor-heavy analytics tend to reside on the user’s server. Images are sent from the cameras to the server, where the analytics software processes them, and are then presented as part of a video management system - the user interface that can control cameras, view real-time images, review earlier footage, and present the results of integrated analytics programs.
Camera manufacturers will often bundle their own management software with their system, but another option is to use open video management software, such as platforms from Seetec or Milestone Systems,
as these tend to work with a wide variety of camera manufacturers and analytics providers - this can make it easier to tailor a system to a user’s specific needs.
Some leading video analytics vendors include Agent Vi, Immersive Labs, Ipsotek, ObjectVideo and SightLogix - although this is a growing sector and new companies are emerging regularly.
This is one of the reasons that using open platform video management software can be beneficial, as often new analytics providers can be integrated into the systems.
What can it do for retailers?
Vy Hoang, sales and marketing vice-president at i3 International, which has provided analytics-based business intelligence solutions for retailers including Urban Outfitters and DKNY, says the ability to deliver real-time reports is a big advantage.
He says: “Say, for example, that you have a display of expensive sunglasses as a shop-in-shop. It might be located adjacent to the cosmetics department or next to a custom jewellery counter. Not only do you want to ensure that visitors to the store do not help themselves to a pair of shades from a loss prevention point of view, you might also want to track and analyse which brands and designs attract most customers, as this could prompt you to rearrange the display.”
Hoang says the way that goods are presented and small changes in the way they are displayed can help to provide a more user-friendly experience for the customer, and can even encourage them to spend more. Use of a retailer’s property in this way can generate immediate impact on the bottom line.
He adds: “You might want to track customers who bought sunglasses around the adjacent shop area to identify which goods they also sampled or bought in cosmetics and jewellery, thus investigating if certain behaviour or movement patterns apply to certain categories of customers.”
Using video analytics software on the footage will provide reports covering view time in front of a display, walking patterns within and around the area, as well as traffic counts on the number of people in the area.
According to Atul Rajput, business development manager at network surveillance camera manufacturer Axis Communications, network video-based footfall counters provide real time updates across the store estate, giving retailers access to data that enables them to plan staff levels to match customer traffic. Making sure there are enough staff in store at peak trading hours will reduce customer dissatisfaction and improve service.
Rajput adds: “Retailers can also track customer behaviour using analytics integrated with other network data sources, such as point of sale transactions, allowing retailers to compare traffic and conversion rates not only between stores, but down to aisle and display location, and monitor the success of promotional campaigns and window displays.”
By counting the number of people in each part of the store, heatmap analytics can visualise ‘hot zones’ - those with the most customer activity - to help maximise in-store promotional campaigns, and identify ‘cold zones’ to determine how store layout changes can improve customer traffic flow.
Data from CCTV can allow retailers to immediately evaluate the impact of floor change layouts on customer flow and sales, by combining mapped traffic patterns with POS data, Rajput says.
It’s not just about numbers either - other data on shopper reactions and specific movements can be gathered. Retailers can log statistics relating to customer dwell time on in-store advertising or digital signage, and track customer flow. Capturing responses to displays and promotions can help retailers to optimise display and marketing strategies and drive sales.
CCTV can even help cut queues, Rajput says. “An integrated network video intelligence platform can generate real-time alerts when queues exceed predefined thresholds,” he says. Staff can be alerted by the system, and open additional tills or replenish stock. It can also help them address sources of common frustration experienced by time-conscious customers, with the aim of improving the shopping experience and hopefully increasing customer retention levels.
Rajput says: “Some smarter queue analytics systems can also integrate with footfall data and, based on the volume of traffic entering the store, can alert retailers of the need to open additional tills proactively before queues start to form.”
Video analytics software can also shine a light on who exactly your shoppers are. Rajput says: “Intelligent analytic applications that capture and analyse demographics, such as age and gender, can also turn video information into valuable intelligence on customer buying patterns and shopping habits of different demographic groups.”
This can be achieved through integrating network cameras with facial recognition analytics, which can determine factors such as the gender or approximate age of the customer looking at a digital signage display.
An additional area where facial recognition is beginning to stoke interest is being able to identify ‘unique’ visitors. The system does not identify any specific individuals, but provides data on the number of unique visitors to a store - useful if a retailer has launched a campaign designed to attract new customers.
Track and trace
Companies such as AMG-Panogenics - manufacturer of the aforementioned 360-degree camera - and Ipsotek have developed ‘track and trace’ capabilities for their products.
David Myers, chief technology officer at AMG-Panogenics, says: “By detecting and tracking motion in pre-set areas, our tracking camera streams can follow people around within the 360-degree view of the camera. We can currently track two individuals simultaneously, and the main intention of the feature is to act as a deterrent to would-be shoplifters by displaying them on the in-store screen and tracking them as if they were being
The live tracking streams of different areas in the store are generally displayed on in-store monitors around the relevant shopping area, together with a message along the lines of “shoplifters will be prosecuted”.
Once a particular area is entered by the shopper, the camera is triggered and the tracking automatically starts.
The customers therefore see themselves on the display screen while browsing and, when they move, the camera follows them.
Ipsotek’s version - Tag and Track - uses a network of cameras to tag and then automatically track an individual. Operators can tag the individual manually in real time, or can program criteria - such as entering a restricted area - which will tag them automatically. The system will also track the tagged individual backwards in time to show where that person has been.
With the rapid pace of change in computing technology, we can expect to see continued significant evolution in analytics capabilities.
While we talk of ‘facial recognition’, what we usually mean is ‘face recognition’ - that is, the system recognises that a face is present, can identify with some accuracy whether that face is male or female and can estimate age. But real facial identification is much more difficult to achieve.
At present it’s much more common for a system to be able to identify and alert an operator to a small number of enrolled faces - say, 10 or 20 people. And while this is undoubtedly useful, it’s likely that if a retailer had only a small number of known suspicious individuals it needed to identify, staff would probably have a fair idea who they were anyway.
But this will change - as processing power improves, so too will the capabilities of facial recognition analytics.
It’s likely this will be pioneered in anti-terror applications first, before making its way to retail.
Another developing area is audio analytics. This is where microphones pick up sound, which is then analysed in real-time (not, generally speaking, recorded - there are privacy rules about this in most places) for signs of aggression.
If something potentially worrying is detected, cameras can be triggered to automatically switch to the sound’s location, and an operator alerted. Although it has been used mainly in prisons up until now, there is potential to develop audio analytics to monitor for non-aggressive speech - for instance, to discover how many customers discussed a display or made positive noises about a product. And there’s definitely a potential use for monitoring for aggressive sound in late-night, single till operator retail stores.
Perhaps some of the most significant developments are being pioneered by companies such as Immersive Labs, an American business that specialises in facial detection software.
They gather detailed, but anonymous, demographic information, assessing gender, age range, attention time, what people glance at, emotions, and even individual customer interests - formal clothing or casual clothing, for instance. Based on this, the company can automatically target advertising on digital displays to fit the customer profile. This can be used in addition to more standard analytics such as monitoring customer paths taken through the store, queue performance, and entry and exit points.
We can expect to see more of this Minority Report-style tailored marketing in the years to come. So what’s the hold up?
There isn’t one, really. A number of big UK retail chains are using video analytics already, though the majority aren’t that keen to talk about it, for reasons of both security and competitive advantage. However, Axis points to the work it has done with Tesco, using surveillance to identify internal fraud and extract POS information.
And although they didn’t want to be identified, a major UK high street retailer is using Axis cameras and Agent Vi analytics for people counting, heatmaps and to measure dwell time.
But according to a recent survey by the Centre for Retail Research, the main barrier for retailers in moving to IP-based systems and the use of video analytics is the need for collaboration between the loss prevention and IT departments.
Professor Joshua Bamfield, from the Centre for Retail Research, says: “This survey has highlighted how retailers are embracing technology and recognise the role that surveillance can play, but there is still some resistance, due to issues with collaboration between departments, that is hampering the adoption of IP surveillance.”
Costs for implementing video analytics will, of course, vary. If you’ve already got an IP-based CCTV system in store, a decent integrator will be able to advise on how best to introduce analytics. An even better opportunity would be if you are considering upgrading from analogue CCTV to network-based - that way, new cameras can be positioned in ideal locations for both surveillance and business intelligence gathering purposes. Again, costs will depend on the scale of the installation, and the existing infrastructure.
Return on investment is, as always, a tricky beast to measure. But if you can use cameras to reduce shrinkage and shoplifting while at the same time producing measurable data that helps to sell more products, video analytics is a good bet in pointing your business in the right direction.