Paul Osborne, director international projects Romaco Laetus, provides a guide to machine vision systems
Pharmaceutical manufacturers must ensure that products leaving the production line are correctly assembled with the desired components. An essential part of the GMP control for consistency is a need for electronic supervision and, perhaps, an additional desire to compile comprehensive statistics to predict when, say, preventative maintenance will be required.
Machine vision is continuing to prove to be a valuable tool for increasing control and efficiency. Not yesterday’s dedicated vision systems though, but rather the new breed of industrial PC based systems and powerful intelligent cameras and sensors that are easier to use, less expensive to supply and maintain and more fully featured than in previous years. To understand the application of these technologies it is useful to understand the vision marketplace which is comprised of two market segment.
Firstly there is the application specific, OEM segment; here the application is well defined by the nature of the production machine and its end product. Blister packaging for tablets and capsules is a good example. The requirements for inspection of completeness of pack, correctness of individual colour and shape, while inspecting for rogue items inside and outside the blister pockets, are attainable at high line speeds with 100% accuracy with today’s PC based systems.
Secondly, there is the so called General-Purpose Machine Vision segment. This market is the most difficult to supply into and generally requires the greatest amount of flexibility of system design and functionality. Here the supply of equipment should be considered as a project, with attached costs that go beyond application and development work and include creation of validation documentation.
How are machine vision systems defined? Broadly speaking they fall into three distinct groups: vision sensors, smart cameras and vision systems.
Vision sensors incorporate the components of a machine vision system, including the camera, digitiser, processor, software, interfaces, optics and illumination, into an extremely compact package. These devices are programmed by a technique known as ‘Show and Go’, where a good sample shape is shown to the device and all subsequent evaluations are based on some type of correlation to the shape of the original model. These products can be used in a variety of applications such as simple print detection and level sensing. The field of view of these devices is often limited to less than 50 by 50mm.
GPMV smart cameras are similar in construction and concept to vision sensors, the main difference being in the larger size of the device, the greater depth of application programming, the power and speed of processing, the field of view and its resolution. Instead of programming in a comprehensive set of operations, users are generally required to develop applications based on library components provided by the manufacturer. While using a library simplifies programming, it also serves to restrict functionality and it may not be possible to realise a desired specification using this technique alone. Additionally, the manufacturer will usually supply little or no validation support.
Application specific smart cameras are the growing trend in the pharmaceutical market. This is when the camera manufacturer decides the complete specification for the camera operation and releases it as a completed device with a harmonised Human Machine Interface for specific applications. This approach is generally more acceptable for the pharmaceutical industry since the main configuration testing and documentation will be done by the manufacturer and he will have prepared the background work for the validation of the device in its final application.
Vision systems are usually designed specifically for inspection tasks such as blister or label inspection. The reason for utilising a complete vision system rather than a smart camera is that a PC-based computer with separate camera will still today give certain advantages. The first is processing power; using a PC-based device enables the manufacturer to keep up with the evolution of the latest, fastest, microprocessors, as they are released. Today we can expect to see the power of industrial PC-based vision systems double every two years. Secondly, the open architecture of the PC system lends itself to the addition of extra processing components. Finally, the PC system comes with relatively low cost and yet very sophisticated operating systems for the use of the vision system designer. Nowadays the use of the Windows XP operation system in its embedded form gives a very robust system indeed. Embedded operating systems mean that the vision system may accidentally lose power, without the chance of the vision systems operation becoming corrupted.
It would not be a complete article on vision system without a consideration of connectivity. For a long time machine vision suppliers have looked to be able to integrate their systems with other devices and with the machine control itself. This is being realised by the adoption of the same communication method that connects office computers together, TCP/IP. This was developed originally to connect a number different networks designed by different suppliers into a network of networks [the Internet]. It was initially successful because it delivered basic services that everyone needed [file transfer, electronic mail and remote log-in].
The latest development in software communication between systems is that of the Extensible Markup Language. XML is not just for web pages, it can be used to store any kind of structured information and to enclose or encapsulate information in order to pass it between different computing systems which would otherwise be unable to communicate. XML allows the flexible development of user-defined document types. It provides a robust, non-proprietary, persistent, and verifiable file format for the storage and transmission of text and data.
Today it is possible to connect the vision system, additional security devices like bar code readers all to a point of central control, which may even be software included in the main machine control panel. Additionally, this interconnectivity means that mechanisms such as 21 CFR part 11 can be realised and maintained for the complete machine security from a central location, as well as enabling sophisticated data transfer between various systems.
Finally, what are the trends in pharmaceutical vision systems for the future? I believe these will be in a variety of directions.
Colour is now standard in the industry for many inspections using vision systems; this technique will spin off into high resolution smart colour cameras
Colour vision system will evolve higher resolution with many millions of pixels, capable of resolving very small objects. This will be achieved by faster communication between camera and vision system and the higher PC processing speeds previously mentioned
Moving out of the visible spectrum, actual vision systems, rather than just spot scanners, will move into the near infra red region from 900 nm to 1700 nm [0.9μm–1.7μm]. This can be used for the physical analysis of substances during secondary manufacture. By obtaining spectra from multiple points on each individual tablet or vial in this NIR region, it is possible to determine a number of parameters relating to the tablet or vial chemical composition and dosage.
Improved control and exchange of information by use of the Extensible Markup Language enabling standardised data transfer between vision systems and their human interface.