FEATURE

Adverts that know what you want

Advertising boards that recognise the consumer and play them tailored adverts were once the thing of Hollywood, but are now becoming a reality thanks to machine vision. Tom Eddershaw reports

Consumer attention and the contents of their wallets have been the battle ground for advertisers since advertising began – but now, by using intelligent imaging techniques, tailored advertising strategies can be applied.

This is increasing the success rate of an advert and also, depending on your point of view, is beneficial to the consumer by identifying fundamental characteristics and selectively displaying content relevant to that potential customer. Features such as approximate age and gender are pulled from images by the system’s algorithms, helping it decide on the customers demographic and which adverts to show.

However, the technology used has been reapplied from the more saturated security market, which raises questions of privacy.

The issues are largely fuelled by negative publicity surrounding this type of advertising strategy. The UK supermarket chain Tesco began using such systems, which prompted a range of mainstream media to compare the technology to the 2002 Tom Cruise blockbuster Minority Report.

In the film recognition software is used to tailor adverts to specific customers. Previous purchases are remembered – as well as the demographic data – and this is used to create a personal advertising experience.

According to Paul Coombes, head of commercial at Argus Global, he has to deal with this comparison to Argus’ own technology and help customers understand that this is not how things are done. Argus is an integrator of this type of intelligent system who has been involved in facial recognition for some time. The company has recently focused on retail applications. Coombes said: ‘Minority Report is actually targeted advertising based on a specific individual. What we do is targeting advertising towards a general type of person. So it’s actually more subliminal and less in your face.’

The fundamental difference is that a typical system doesn’t store any personal information about a person and it normally doesn’t store any images. The technology could be linked to security systems, but this means it has to abide by the same legislation as the CCTV operations.

When a shop is using a CCTV system, it would have to make it known that the system was in use. There is no specific legislation for tailored advertising systems, but Argus recommends that its customers put up signs even if they are just using digital signage technology. Coombes said: ‘From our point of view we like to address the privacy misconception issues head on and say “actually, it’s a misconception that this is Minority Report, and the privacy issues aren’t what you think they are.” It’s only by getting people to challenge us that we get the opportunity to inform people. We have a code of conduct that we give to retailers and believe from our industry experience that this is the best method of practice to improve consumer approval.’

He continued: ‘I often liken it to what you see in supermarkets. They may have a stand of samples, perhaps it may be a baby-related product, and so the salespeople would actively target people with children in pushchairs. Or the perfume and aftershave salespeople in duty free at airports; the perfume seller won’t target men, it’s more likely to be the aftershave sprayed in your face. They will know the target market that they want to aim at with their advertising which is what we are doing but in a digital way.

‘The only difference from a privacy point of view is that you are more likely to have more trigger adverts that are relevant to you. That’s as far as it goes really; you still have the choice to buy or not to buy, but the point is, if you are stood in a queue and there is a cycle of ten adverts going on, they could be ten adverts that are more likely to be relevant to you.’

‘We often have to talk to people about the misconceptions of the technology. They are assuming it’s Minority Report type stuff, but if Hollywood actually reported on what today’s technology was doing, it would be a much duller story.’

Today’s technology looks at a person’s face once an image is captured by a camera. Using patented algorithms the system then decides how old the person is, their gender and other defining characteristics that can be used to monitor the kinds of people passing the camera, or simply to select relevant adverts.

Argus uses two way communications between the camera and the rest of the system. The cameras are providing roughly 15 frames per second, and the system is set up to cater for around 25 faces in a frame. Once a face has been detected, the system enhances the face image and only when the quality of the image is maximised does the system apply the algorithm to determine age, gender and other characteristics.

Coombes said: ‘We don’t have a standard off-the-shelf system. For example, there is a fashion retail store working with us at the moment that wants a camera at the entrance on its own, with a digital advertising board within the store. Or it could be mounted with the advertising board because the form factor for machine vision cameras is pretty small. But as long as the systems are linked by a network, the processing could be done anywhere within the building.’

This means that multiple cameras could be used to decide the rough demographic of the clientele in a shop at any given moment and select adverts based on what is relevant to the collective. Or, a personalised version focused on the person stood in front of the system could be created, optimising advertising on a case-by-case basis.

Further to this, the data could be used just for the shop owners to understand the type of people that are entering their shop, with a report being printed out at the end of the day breaking down the selected data.

Argus Global has been involved with biometric system integration since 1997 and its systems have been deployed worldwide in high security facilities such as prisons and police forces where the identification of people is paramount. Coombes said: ‘Our reputation and company brand comes from our successes in high security markets where misidentifying people is simply not an option. When entering the vertical market of retail analytics, which is more about supplying data about consumers in an anonymous manner, we applied the same approach as we have always done and that is to ensure that we work with best of breed providers.’ In the case of face recognition, Argus chose to work closely with German-based provider, Cognitec Systems.

Face recognition has been used in these types of high security applications for many years. But this differs from the newer directed advertising field because instead of extracting an image from a passport and comparing it to a live image of a person for airport security, Coombes clarified that, while the same technology is used, none of the data is stored and it is completely anonymous.

It has taken time to bring face recognition technology into this new field. Coombes said: ‘It’s mainly down to the manufacturers who were working within a space, border patrol in particular, which experienced sudden and incredible growth. As soon as the technology was available, and disasters such as 9/11 and the other global terrorism events happened, governments invested heavily in biometric security within border control.’

However, as the market became saturated because of more companies entering the field and more competitive because most people that needed the technology already had it, manufacturers began looking at other areas and other industry verticals that would provide additional growth.

‘There is a difference between what Cognitec do that separates it from much of the competition,’ Coombes stated. ‘There are lots of people that are entering the marketplace that you’ll see using IP [internet protocol]-based cameras that are used to feed information through to the software. Let’s say for argument’s sake that one algorithm is as good as the next, the next differential is the quality of the information you are providing to that algorithm. As a rough rule of thumb: poor quality data in equals poor quality data out.

‘This is one of the reasons why machine vision cameras, such as those manufactured by IDS, have been integrated with the Cognitec systems. When you actually get down to the fine detail of imaging, there is often a big difference between standard IP-based CCTV cameras used for recording wide areas, and can just about identify people, whereas our system requires the ability to analyse large amounts of data points extracted from a person’s face image.’

He also said that CCTV cameras don’t tend to have fixed focal points; PTZ cameras give the flexibility to focus on different points within a store or to focus on different areas whereas the machine vision cameras with fixed optics can be optimised to give the best performance from people walking in front of a camera. Argus also prefers to use a global shutter as there is reduced blur.

Coombes concluded: ‘We believe this is an exciting time for manufacturers, integrators, retailers, and consumers alike. Retailers can expect to be able to receive valuable, accurate, real-time data about their customers, much of which is currently only available when consumers make purchases linked to their loyalty cards. This anonymous information is gathered without the need for consumers to link their purchasing habits to any other aspect of their personal data. So when you think about it closely, we give retailers access to much more of our personal data every time we swipe a loyalty card to get a few extra reward points.’

Engaging the viewer

Intelligent advertising and camera technology is enabling not only selective media campaigns, but deeper engagement of the audience as well. London-based 18 Feet & Rising was tasked with creating an advertising campaign for the car manufacturer Škoda with the platform, ‘the car that will steal your attention’. The result was a split screen video of Škoda’s cars in different environments, but by using eye tracking technology from the technology company B-Reel and borrowing the user’s webcam, viewers could find out which of the videos attracted their attention.

When speaking to Imaging and Machine Vision Europe, an 18 Feet & Rising spokesperson said: ‘Our brief was to activate that [platform] in Škoda’s digital channels across Europe. Deeper engagement is critical online; otherwise you stand to miss the opportunity to spend more time with the consumer. The ambition to do something new was always there. We wanted to develop something that the public had never experienced, something they could genuinely engage with while keeping the brand, and the product, at heart.’

Explaining the process, 18 Feet said: ‘The creative answer was an online experience that connected consumers both physically and technologically with the product, for a minimum of two minutes. The biggest hurdle was implementing reliable eye-tracking tech that wouldn’t require specialist hardware. We searched far and wide, met with people from all walks of life but kept coming up short. Many people claimed to be able to eye-track but in reality, under testing conditions, the existing technology was very poor. It took approximately five months to get the tech up to scratch. The pace in innovation was remarkable and the final tech was very reliable.’

In terms of implementation, all creative responsibilities including film production, were developed by 18 Feet & Rising in London working with production companies in the Czech Rep and Portugal. The bespoke technology and builds were handled by B-Reel, in collaboration with Škoda’s CMS system in order to enable implementation.

18 Feet said the project was a success. On day one of the central European site’s launch, unsupported by any media spend, the site received three times as many unique visits as their previous campaign – which was supported with media – received in two months.

‘We will most certainly use this technology again. There are numerous creative opportunities for it and as reliability continues to improve, so does the potential for the user experience, given the ability for greater complexity in application.’

About the author

Tom Eddershaw is a technical writer for Electro Optics, Imaging & Machine Vision Europe, and Laser Systems Europe.

You can contact him on tom.eddershaw@europascience.com or on +44 (0) 1223 275 478.

 Find us on Twitter at @ElectroOptics, @IMVEurope, @LaserSystemsMag and @ESTomEddershaw.

 

Other tags: 
Feature

Damir Dolar, managing director of Croatian firm Smartek Vision, spoke at the European Machine Vision Association’s business conference in Dubrovnik, Croatia, in June. Here, he gives his view on the European industrial imaging market

Feature

Matthew Dale discovers that polarised imaging could soon see a surge in uptake in industrial inspection, thanks to a new sensor from Sony

Feature

Active Silicon turns 30 in September. Greg Blackman speaks to founder Colin Pearce (below) on building the company and the importance of FPGA technology

Feature

Greg Blackman on the deep learning software tools now being deployed in industrial imaging

Feature

Keely Portway looks at some of the miniature imaging technology used in endoscopes

Analysis and opinion