Ever more stringent quality demands increasingly call for 100 % inspection. Individualised low-volume production, miniaturised components, high production speeds, or monotonous testing procedures are likewise issues prompting greater use of industrial image processing in quality assurance. Until fairly recently a topic solely for experts, machine vision has meanwhile developed into a readily and economically applicable versatile technology thanks to advances in computers and cameras and thanks also to standardisation and a significantly improved price/performance ratio.
We may well be unaware of the extent to which we use machine vision and industrial image processing on a daily basis. They make our everyday lives so much easier. For example, the camera of a smartphone not only takes photos. It can do much more: Data captured by scanning business cards can be directly entered into a smart phone’s address book. Thanks to intelligent image processing, photographed sites of interest are recognised and our smartphone can even understand scanned words and translate them into the desired language. In industry, image processing – machine vision – is used for guiding and controlling robots, for process control by identification of barcodes, for tracking & tracing on the basis of barcodes, or as non-contact testing procedures for product quality assurance. Cameras use various methods to record images, which are processed to yield metrological data and quality assessments. Machine vision can be used in quality assurance for surface inspection, for shape and dimensional inspection, for position or object recognition, for layer thickness measurement, or for completeness checking. These tasks may vary widely in their degree of difficulty.
Increasing Levels of Difficulty
Completeness checks and so-called pick-and-place tasks number among the simpler challenges. Here it is necessary to compare certain patterns with a reference. Counting, searching, and comparing are somewhat more complex, as is reading of codes or characters. Position or angular position recognition of individual articles often serves for control of robots. Yet all these tasks do not amount to quality assurance of a product. The tasks involved in quality assurance are of a more complex nature: Inspection of surfaces, of dimensions, and of colours is more difficult for machine vision.
Image Processing Is Here to Stay
Today, industrial image processing has already secured a firm place in the plastics industry – and the number of applications is steadily increasing. According to the German Engineering Federation (VDMA), Frankfurt, deliveries of machine vision equipment to the rubber and plastics processing industry increased significantly (by 33 %) in 2011, after a period of steady decline between 2007 and 2009. Increased volumes of plastics packaging as well as plastic products for the medical device, pharmaceutical, and cosmetics industries translate into greater market potential for industrial image processing. The plastics industry accounted for some six percent of sales of image-processing systems in 2011 and also in 2010. This is unlikely to change much in the near future, according to Patrick Schwarzkopf, Director of the VDMA Machine Vision Group.
Market data issued by the European Machine Vision Association (EMVA), Barcelona, Spain, reveal a similar picture. Gabriele Jansen, Member of the EMVA Executive Committee and CEO of Vision Ventures, Heppenheim, Germany, offers the following interpretation of the numbers: „Although just a small portion of the total annual sales by European suppliers is directly assigned to the rubber & plastics sector – i.e. about 4 percent – almost 14 percent of total sales are to containers and packaging, another six percent to pharmaceuticals and cosmetics, and almost 3 percent to medical devices.“ These are all areas in which plastics processing has an important stake. „Plastics processing also plays a significant role in a number of other sectors, such as the automotive industry and its suppliers,“ says Jansen.
In Support of Sustainability and Production Efficiency
The inherent potential of industrial image processing suggests that it will be the quality assurance technique of choice for production methods of the future – precisely for components and products made of plastics. Current debates concerning production efficiency and sustainability indicate that in the future we shall be forced to derive maximum benefit from minimum possible consumption of materials and energy. High reject rates are incompatible with this trend. In many areas there is also a clear shift towards individualized products and small production runs. Today, impeccable quality, frequently requiring 100 % inspection, is almost taken for granted by some users. Examples are to be found in the automotive industry or in medical technology. For the plastics processor this means that the parameters of every single component, every centimetre of a film or strand, and each individual packaging will have to be checked. This task extends far beyond the capabilities of the human eye.
When is Image Processing the Appropriate Choice?
However great the enthusiasm for the method may be, Jansen warns against using a sledgehammer to crack a nut: „Image processing can only be used as a method of quality control if the quality of a component can be presented visually. If this is the case, it should nevertheless be examined whether other test methods are possible, and what the pros and cons of the various methods may be.“ She continues: „If quality control is possible with a simple optical sensor then it should be carried out with such a sensor and not with a camera-based system. The simplest solution is usually the best.“
The quality of the component must be visually recognisable. Internal structures or defects cannot be made visible. Automated test methods – such as industrial image processing – can also be useful if a high reject rate is encountered because the product is not tested until the end of the production process, or if the inspection tasks are too complex for the human eye or too monotonous for the human brain. Inconsistencies in performance can turn humans into uncertainty factors. This warrants special consideration in the case of safety-relevant parts, which often require 100 % inspection. Schwarzkopf adds: „Traceability may likewise be a relevant factor, and miniaturisation is also an important aspect. And very fast processes, such as film production, and high production volumes are clearly also good arguments favouring use of industrial image processing.“
Fuel filler doors of vehicles are examined, for example, for surface defects by optical inspection before they are painted. Jansen mentions other examples: „For web materials such as plastic sheets and films, line-scan cameras detect minimal surface defects at high speed across the entire web width in 100 % inspection. Image-processing systems monitor charging of and discharging from processing machines by 2D to 3D position determination of the parts. After injection moulding, image processing is used for 100 % quality control of the moulded parts, possibly also with thermographic cameras. Surface inspection of mobile phone shells, dimensional inspection of precision parts, and also print quality inspection of plastic bottles as well as track & trace on the basis of 1D- and 2D-codes are all examples of successful industrial image processing in the plastics processing industry.“ Image processing can even be used in maintenance and hence in preventive quality assurance. Thus optical scanners permitting automated 3D digitisation can be integrated into a machine to monitor the wear of a forming tool, thus ensuring its timely replacement.
Take Advice and Learn from Experience
Sound advice prior to the first use of industrial image processing is just as important as careful planning of the overall project. As with any project, the first step is to draw up a detailed requirement specification sheet as a basis for a functional specification to be developed jointly with the image-processing system provider. „Here it is particularly important to bear in mind that the image-processing system relies on visual information, meaning that the optical stability – or possible instability – of the product and process has to be considered in the specifications“, Jansen emphasises.
Care should be exercised in the choice of supplier; the criteria applied should extend beyond mere price considerations. For example, experience, references, breadth and depth of any offer, available resources, and staff accessibility all play an important role in determining the success of a project. Yet not every image-processing application is necessarily a large-scale or complex project. Standard solutions exist for a multitude of tasks, including components which the user can integrate into the system. „However, this requires a certain amount of experience which a first-time user will need to acquire“, says Jansen as a word of caution against an overzealous buy-plug-and-play approach.
Image-processing systems are frequently already included in the planning of production lines or a machine. That has advantages for the mechanical, electrical, and control-engineering integration of the system, and also for its ease of operation. However, image-processing technology is advancing very rapidly and thanks to a constantly improving price/performance ratio it is finding more and more applications. That is why retrofitting of existing production lines continues. And once a new application has proved itself in practice and is seen to be cost-effective, it will be included as standard in the planning of similar plant.
Limits of Technology
For an image-based process the quality-relevant parameters must be of a visual nature – or must be capable of being visualised. Internal structures and also internal defects – such as inclusions – can hardly be recognised. „Although tentative steps have been made to utilise computer tomography as imaging technology in a production environment, a number of technical challenges still have to be overcome“, Jansen indicates. So-called ‘uncooperative optical conditions‘ also prove problematic. These include pronounced fluctuations in contrast. The influence of daylight in the test area can be reduced through suitable light protection measures. „Where there is a lack of contrast, for example in the case of the characters embossed on the walls of rubber tyres, attempts have been made to utilise 3D image processing as a method of ‘contrast generation‘: Height information instead of greyscale information“, Jansen explains. Schwarzkopf adds: „Combination with other techniques – such as thermography – can open up new applications. This is particularly true in the area of injection moulding.“
Rapid Advances in Technology
Just a few years ago the rate of measurement was a topic of discussion in relation to the use of industrial image processing for quality assurance. Today it is hardly ever mentioned. On the contrary: New technical advances – including 3D image processing, multispectral technologies, image sensors with ever-higher resolution, speed, sensitivity or dynamic properties, as well as the cameras based thereon with their corresponding interfaces – are opening up more and more applications. In parallel with developments in computer technology, the price/performance ratio is becoming ever more favourable, components are becoming smaller, and handling simpler. Schwarzkopf comments: „Increasing standardisation also facilitates application, thus favouring market penetration.“ This is where the trade associations – EMVA and VDMA – can make a significant contribution.
- Verwandte Artikel