Everyone has heard horror stories about reaching the dregs of a bottle of drink before finding a mouse had somehow crawled inside, died, and been embalmed in alcohol some months before.
One hopes that in the modern age this should prove impossible, thanks to advanced vision sensors such as those built by Banner Engineering. It recently provided a customised system (pictured above) to detect foreign bodies in drinks bottles for an unnamed supplier. Banner’s system is primarily designed to detect vent tubes, used to insert air and liquid into the bottles, that may have rattled loose during the high-throughput process. In the past, the whole of the line would need to be quarantined, as the tube could potentially choke consumers.
The system detects the objects using a very bright backlight that creates a high-contrast silhouette of any foreign objects, making it very easy for the software to analyse and pick them up, even at very high speeds.
‘Even rates of 800 bottles per minute are no problem if you use enough light,’ says Jeff Schmitz, the corporate business manager for machine vision sensors. ‘The name of the game is to create the highest contrast you can, with a dark silhouette against blinding light.’ An actuator may then pick up the bottle, or the whole line might have to be stopped. Clearly such systems need to be very robust, especially where liquids are concerned. Schmitz describes ‘an elegant enclosure’, where the camera, illumination, and optics are built into one package, specially water-sealed to prevent damage. The control device is in a separate cabinet, with additional input and output functions.
‘The optics and lighting are very critical,’ explains Schmitz. ‘It would be very easy to defeat a machine vision system, so it is a challenge to make it work robustly.’ It is for this reason that Schmitz believes machine vision may not have been as widely accepted in these applications as the technology warrants, with just a few past failures discouraging manufacturers from using it widely.
However, improved technology, faster processing and a rapid drop in prices – Schmitz estimates these inspection systems may now cost 10 times less than they were five years ago – has meant many more drinks manufacturers are now willing to take the risk. A suitable example of a product that has been optimised to the bare minimum requirements for such operations would be the Cognex Checker, which is now available for just £1,000 as opposed to more like £15,000 some years ago.
Checker has recently been used by US company Meridian to verify whether bottles had been filled and correctly capped at a rate of 375 per minute. It detects seven different features using around five images of each bottle to reduce false alerts. The Checker is perhaps most suitable for checking labels on bottles at high speed, for the alignment, date code and language. Clearly, optical character recognition and pattern recognition are necessary features of the software, but few other capabilities have been left on the DSP (digital signal processor – the ‘CPU’ of a smart camera), allowing it to drop to a price more affordable for most production lines.
It may seem unimportant, but the label alignment is actually a finishing touch that could make or break a product. Jeff Schmitz argues that when selling low-price drinks like sparkling water or squash, there is little differentiating the products apart from the marketing, so presentation is everything. At the other end Declan O’Dea, regional sales manager of Cognex for North Europe, believes that a polished finish is especially important for producers of spirits and other high-value products, even to the extent of rejecting a whole crate if the label is just half a millimetre out.
In addition to attaining this precision, Cognex has also spent a lot on R&D to find the best way the system can be set up and used by engineers. ‘The main issue is ease of use,’ says O’Dea. ‘It’s something that’s prevented vision systems gaining fully blown acceptance in the past.’
One of Cognex’s newest products, OmniView, available since mid 2006, is ideally suited for bottle inspection. Rather than spinning the bottle around, which would slow down the process and could possibly damage the product, the system uses four cameras placed around the bottle to capture each side. Software then pieces together the four separate images successfully, providing a full 360° inspection at rates of more than 1,000 bottles per minute. However, currently this kind of processing needs a PC platform, so a fully contained smart camera system is not available yet. If that seems advanced, Sick IVP has produced a 3D imager that can detect faults as small as 0.2mm in can lids, using time-of-flight laser measurement. Once the cans are filled and sealed, they are sterilised in an oven and heated to 90°C. However, if the lids are open just a tiny bit, they will explode, destroying a lot of the equipment and potentially costing the company more than £30,000. Clearly even very expensive imaging systems would prove to be a good return on investment.
This kind of accuracy would only be possible using 3D imaging. Anders Murhed, manager of business development, explains the other advantages: ‘Everything is contained within the unit, including the illumination – the laser, the lens and a very narrow band pass filter that only lets through the laser light.’ In particular, this means that the colour or pattern of the bottles being inspected makes no difference, whereas with 2D systems you would need a good contrast between the object and the conveyor belt: something Coca Cola has taken advantage of, when using it to check the lids have been tightened on large bottles of coke, that otherwise could come undone during transport.
Sick’s 3D vision is also used in the manufacturing of bottles, to check the hot plastic plates that will then be blown into the final shape. The plate must be the right size, so that there is enough plastic to make the bottle, and it can not be bent otherwise it will break in the machine. This high-speed 3D vision, through a smart camera, was only developed a year and a half years ago. It seems the advent of the smart camera has been instrumental in reducing the price and allowing a number of these applications, even 3D, to finally open up and blossom. For the future, Sick plans to create a wider range of cameras, with different fields of view and resolutions, to serve even more applications.
‘By packing it into a smart camera, we’ve made a very economical solution,’ says Murhed.
Schmitz agrees that the price and availability of these cameras is finally making them acceptable to bottle manufacturers. ‘I don’t think this trend is stopping. It just continues, making all of us stronger.’
The software libraries that provide the analysis are just as important as the vision equipment used to take the images. German beverage company BBull uses MVTec’s Halcon software library to bottle Coca Cola and Korean and Irish beer.
BBull’s systems check all areas of the bottles labels, for fit, expiry data, captions, logos, and scratches and creases. It also reads bar codes and data matrix codes, no matter what position it is placed or how the bottle has been rotated.
One of the key features that attracted BBull was its ability to automatically analyse four separate cameras in parallel, even at high speed.
‘Without Halcon, our project with this high performance would have been impossible,’ says Georg Krauss, CTO for control systems at BBull.
BBull's system analyses a Coca Cola bottle
