The human touch

Biometric specialist Aurora has been supplying face recognition systems to construction firms since 1998. More than 300 sites in the UK and the Middle East now use SiteRec, its access and time attendance recording system.

In its simplest form, the system controls access using a reader that takes measurements of someone’s face and checks them against a stored set of images. It then allows or denies access through a turnstile. But Aurora’s specially developed software widens its applications to produce timesheets, special reports and allows safety or other messages to be delivered to workers on entrance to site.

Aurora has, in the last year, introduced vein readers which take up much less space than the face recognition set-up. One customer is using this version on two small projects worth around £1m each where workers were fiddling the clocking on system. The technology is saving that client £1,400 a week.

Biometrics in a nutshell

Although it has sci-fi connotations, there’s nothing new about the principle of biometrics. Think fingerprints. Biometrics uses a unique human identifier such as face shape, palm geometry or iris patterns.

Aurora chose face recognition and now vein recognition, which reads the vein patterns on the back of a hand, because after extensive research they seemed best suited to construction. Fingerprints had too many connotations with the law, while iris recognition was rejected because workers can worry about the idea of lasers damaging their eyes, although this is not the case. And palm readers need cleaning too often on a construction site. The other advantage face recognition has is that it’s “human readable”.

You might wonder if face shapes are that unique, but they are. SiteRec takes a series of measurements, for example nose length and distance between eyes, remeasuring each time to check against its base information. It lets less than 1% of imposters, known as “false accepts” through. In practice this isn’t a problem since workers must first enter a personal identification number.

A more frequent problem, although it happens only 1% of the time, is that the system will incorrectly bar someone, known as a “false reject”. This happens because the system is based on a 2D image and if a person turns their head, or the lighting is different, the image changes. To overcome the problem Aurora supplies camera, readers, turnstile and computer equipment in a steel cabin so that the lighting can be controlled.

In future, poor lighting will be less of an issue as Aurora is developing a new camera which will use an infra-red flash.

What if someone grows a bushy beard? The system holds five different images of someone’s face to check against and it can be set up so that after every 10 clock-ins it replaces one of the photos with an up-to-date image.

If someone is denied access, the system stores that person’s photo. This allows potential intruders to be recorded or gives clues as to why a legitimate user didn’t get through (for example if they weren’t looking straight ahead).

Taking and storing the original image is a straightforward task which site security personnel usually perform. Whoever does inductions on site can take down other information which could include CSCS card details and national insurance numbers. SiteRec can also store documents electronically, including signed forms.

Ongoing development

Aurora buys in the biometric technology, but develops all its software in-house. This means that customer requirements, particularly from its first and biggest client Laing O’Rourke, have driven changes and additions.

As well as face details, SiteRec stores information such as when workers clock in and out and it can use that, and other stored data, to generate up to 125 reports. For example, a foreman could ask for a printout of all the joiners on site, with photos alongside.

Laing O’Rourke uses the software to generate time sheets which are then linked to its payroll, which saves time and therefore money.

At the moment databases are site-specific, but Aurora is developing its software so that it works from a central database that holds employee information.

Customers can also choose to go a step further with a PDA-based system for foremen to record more specifically who’s doing what and when. If a group of joiners are carrying out a few different tasks in one day, the foreman can download their details into his PDA and record when they move. This allows for more accurate resourcing on future jobs.

Another useful feature is the ability to display messages to individuals, groups or everyone (see box). These appear when a worker is clocking on and they must press a button to say they have read the message before the turnstile will let them through.

In the same vein

Smaller customers may not have space for the Aurora cabin, which is where the vein reader comes into its own. These are made in Korea where the technology has been used extensively for years, although it is relatively new to the UK. Aurora has been working on it for around a year.

Vein recognition uses infra-red to read the unique vein pattern on the back of someone’s hand. All the same software as for face recognition can be used, and by using a webcam to take a photo at the induction stage, reports with photos can be produced. Vein recognition is statistically more accurate than face recognition.

There can, of course, be objections from workers and data protection laws have to be carefully considered. Aurora allows three levels of password protection. This means that, for example, a foreman may only see names and attendance details whereas the site manager may have access to all the information.

There can be benefits too for workers: the system records exactly what time they have worked and can transfer it to payroll, so there’s no hanging around for overtime payments. The data can be sent quickly and automatically via GPRS, which also removes admin time on sites.