The colour of money

Tom Eddershaw finds that, with new polymer banknotes entering circulation in various countries, colour imaging systems that offer extended sensitivity into the infrared and ultraviolet are needed to inspect the printed money

The Bank of England announced on 13 March that Innovia Security will supply Guardian polymer substrate for the next generation of £5 and £10 notes, which will come into circulation in the UK around 2016. The change will mean that the UK joins a list of 23 countries that, since 1988, have turned to Innovia’s polymer alternative to paper bank notes.

One of the reasons for the switch is to improve security and deter counterfeiting. Paradoxically, this makes it more difficult to inspect the notes as they are being printed in order to ensure that each one is of the highest standard. The inspection of bank notes is carried out not just in colour in the visible spectrum but increasingly in the UV and NIR as well. The techniques can be applied to other printing processes where security is an issue; not just banknotes, but lottery tickets for example. With more than 35 billion Innovia notes printed over the past 26 years, as well as ever more ambitious security features being added, there is little margin for error in the inspection process.

National banks are driving towards plastic substrate banknotes for a variety of reasons. The polymer notes are up to five times more durable than traditional substrates; they are waterproof; more environmentally friendly over time; carry fewer bacteria than paper; and are capable of a variety of new security features that are otherwise very difficult to replicate. In fact, the announcement of the Bank of England contract came the day after the Bank of Mexico reported a 68 per cent drop in counterfeit versions of its polymer 50 peso notes in 2013.

Counterfeiting is a big concern for national banks so they are constantly looking to find new ways to ensure currencies cannot be duplicated. These additional features however, mean even more is asked of colour camera manufacturers to ensure that the wide array of security features can be successfully monitored and inspected.

Mark Williamson, director of corporate market development at Stemmer Imaging, commented: ‘Bank note inspection is one of the most demanding tasks that these [inspection systems] have to cope with. The whole idea of the security print is to have many security features in both the paper and the print to make reproduction almost impossible. Many of these features are not made public to make copying harder, so it is important to validate that these security features are actually present and within specification.’

Accurate colour imaging is needed in many printing applications in addition to banknotes, where security is important, as Teledyne Dalsa’s Mark Butler, team manager for product management and marketing, pointed out: ‘Banknote inspection is a solid application, but package inspection is also a large market. Then there are lottery tickets, driving licences, and also standard document scanning such as the bill papers that you get in the mail; these have probably been scanned by our cameras.’

Changing world

But it’s an evolving market, as Butler pointed out: ‘We are getting a lot of requests from our customers to shift into UV or NIR spectrums as well as red green blue (RGB) and with increasing security requirements – it has increased even more. When our Piranha Colour camera was released in 2005, the colour market wasn’t very big and people were just getting used to RGB space. But once they got through that, and started to understand it, over the past four or five years there has been more and more demand for NIR and UV.’

But this shift poses a problem, Butler added: ‘These spectrums require different colour filters and this becomes difficult from our perspective as a camera producer, as red, green and blue filters are transparent to NIR light. You have to block certain light from getting through different filters while still allowing the light to reach the correct sensor. This is something in our R&D space at the moment. I imagine by the end of the year, you should start to see this type of thing on the market.’

Other manufacturers have tried to overcome the issues with NIR by using all three RGB filters on top of one another. This is still transparent to NIR light and allows it through to the detector while stopping the RGB. However, NIR light is still picked up by the RGB sensors so a subtraction of content was carried out using the value obtained from the NIR filter.

While NIR is predominantly for security, Butler added: ‘The demand for UV is increasing because the things people are inspecting are getting smaller and smaller. On a per pixel basis, object size is now getting to around 5µm. Once you get that small, you start needing blue wavelengths in order to see what you’re trying to find.’

There are plenty of security markers that can make use of the visible spectrum, such as the detail of an image, watermarks, or holograms, and so the system must check for variation here as well. Very small changes in wavelength can be observed across this spectrum by the sensors, which can typically measure 24-bit colour (broken down into each of the RGB channels) which translates into over 16 million colours. Williamson said: ‘Technically you could detect very slight variations within that, however in the real world things will change. You have natural variation in the materials and the manufacturing process. Most customers want to be measuring this variation but if the system was set up to reject the smallest variations you would have a lot of discarded product. The tolerances of the system must be set up to accept small fluctuations.’ He continued: ‘The same product in the morning and the same product in the afternoon could have a different colour value. This is because the illumination source would have heated up, the camera would have heated up, and that can define a massive difference to the entire system. If it had to be really stable, you would have to cool the entire system to a constant temperature and that doesn’t happen because it would be too expensive. Even in banknote manufacture, it is not done to that kind of level.’

Making a mark

Williamson said the characteristics of other security features, such as holograms or watermarks, also pose a problem for the system manufacturers. Watermarks are ‘grown’ and they are not always the exact same shape, causing a difference in colour and shape.

Holograms are very hard to replicate and are designed to look different from different lighting angles. This is a problem for the algorithms that process the images taken from the cameras: if the lighting is at all different, the software may not recognise the feature. The lighting has to be constant, and there have to be tolerances built into the system, Williamson said, and ‘the system must learn the variation – “that’s my model of variation, now tell me if it’s different”.’ So, colour imaging is not just down to the camera. The system typically consists of a sensor, a lens or optics of some kind, lighting, and a processing unit that has image comparing software. The system requires well developed software and high computational power to compare quickly the captured image to the control image.

Björn Weber, product manager at Basler, said: ‘The camera is taking the best possible image in the best possible way with the best possible fidelity. Then the software takes over from there. The system integrators install software libraries that have algorithms, so if a feature is missing it would show up when compared with the control sample.’

Butler said: ‘For a while, separate monochrome images were used for inspection and a colour image was sent to the operator as it was easier on the eye. Now they are able to inspect in colour and feed the raw data out to the controller. Colour imaging can be handled a lot easier now. That’s another reason colour is penetrating the market so successfully, it needed the computational power to improve. Now it’s a fairly straightforward process of doing all three colours at once and combining them into an image. The cameras were more advanced than the software to a degree, but it is all caught up.’

The web of printed substrate is fed under the camera, an image is taken and sent to the computer and processed. If there is an issue, the alarm sounds and that section of the web is taken for further inspection. This all has to be reliable enough to ensure that the final print is accurate; and done at a rate that satisfies the customer’s productivity requirements.

The latest monochrome cameras can operate at a greater frequency than the machines the web is travelling on. Williamson said: ‘For monochrome cameras, the latest camera technology can go up to 280kHz, which is beyond the typical speeds of the web. For colour imaging, the speeds tends to peak around 100kHz.’

A colourful history

For decades cameras have been used for in-line inspection of web printing; however, the original inspections used monochrome cameras. Colour cameras are being relied on more and more for the entire inspection process, instead of just being used for the controller’s copy. Colour allows for a wider array of defects to be spotted and better differentiation between spectral values. For example, with the monochrome cameras, blue wavelengths can be mistaken for green wavelengths if they possess similar intensities.

This could be disastrous for a marketing campaign, for instance. When looking to the history of in-line colour inspection, Teledyne Dalsa’s Butler said: ‘It started with the tobacco industry. They were running out of avenues for marketing and the only thing they had left was how good their packaging looked. This meant they needed perfect prints on their packaging, which meant 100 per cent print inspection of the entire image, instead of just performing spot-checks. This has now spread throughout all labelled or packaged products.’

Teledyne Dalsa’s first colour image camera was released around 14 years ago and was a three chip prism-based system. This dichroic beam splitter separated the light into red, blue, and green wavelengths using optical interference. The image quality produced from layering the three colours upon one another was very high but the product struggled in the market due to practical difficulties. These prism-based cameras are very sensitive to vibrations of the system and keeping the sensors aligned with the diffracted light proved to be quite challenging when used on a fast moving system.  

In 2005 Dalsa introduced the Piranha Colour, which proved much more successful. Butler said: ‘Most customers moved towards this [trilinear camera] even though the overall colour fidelity was reduced. We found that once they started to use the trilinear, the three chip high quality wasn’t required and the trilinear was more cost-effective too.’ The trilinear camera uses pigments, which allow only one of the three colour channels through, coated directly onto the silicon sensors. This means the systems are smaller, require less expensive optics, and still capture a high-quality image.

Last year the Piranha4 8k bilinear camera, which runs at a much higher speed, was released. Only having two colours per pixel, it is a much more compact camera and it is cheaper but the colour fidelity is lower. The camera uses a RBRB pattern on one line, next to a solid green line. Butler explained: ‘We chose a RBRB/G instead of a Bayer pattern because it reduces crosstalk between colours. This can happen when you have colour filters on pixels side by side. Red and blue bleed into each other.’

Somewhere over the rainbow

For further developments in both the security features of banknotes – and other web printed products – and of the cameras used to inspect them, close interaction between the customer and the manufacturer is crucial. ‘When we have a new product development it is our job to inform the note manufacturers of the new capabilities,’ said Williamson, ‘but it is also their job to keep us aware of their demands so we produce the right products.’ Ultimately it is the demand from the customer that is driving the manufacturers to broaden their products’ capabilities.

There is also still growing room in the more traditional RGB colour channels capabilities. Monochrome cameras can still operate faster than colour cameras but this is a gap that is closing. Butler explained: ‘The faster you can go, the better; while still maintaining image quality, the computational power will allow more people to consider colour instead of monochrome. The cameras and the software are likely to move forward hand in hand for a while.’

However, Butler warned: ‘You might now find that the mechanical movement of the system is the next thing to slow things down. Usually what happens when you hit that sort of bottleneck, is that progress will remain flat for a while until someone is able to greatly increase the speed, then we’ll see a jump in further advancements.'


Rob Ashwell looks at the advances being made in scientific image sensors and how these are furthering our understanding of the brain


Shigeki Masumura, president and CEO of Machine Vision Lighting, and winner of the 2016 Vision Award


Recycling is becoming big business, and hyperspectral cameras are exactly the right tool to help sort the growing volumes of waste, as Matthew Dale discovers


Greg Blackman gets to the bottom of Industry 4.0, and asks how machine vision will fit into the factories of the future


Smart Vision Lights has built its company, originally from designing custom illumination, as Greg Blackman finds out