CitationDownload as .RIS
Emerald Group Publishing Limited
Copyright © 1998, MCB UP Limited
Vision for assembly automation
Vision for assembly automation
Don Braggins is a Consultant with Machine Vision Systems Consultancy, High Street, Meldreth, Royston, UK.
Keywords Assembly, Machine vision, Robots
Twelve years ago the Rover 800 series was launched, with windscreens and rear windows automatically assembled on to the body by vision-guided robots. This was supposed to be the beginning of a revolution in assembly, with vision allowing robots to work as if they were human and able to cope with misplacement of components and part-assembled products. It does not seem to have happened that way the windscreens are no longer handled by robots, and the main application of vision stays firmly in the inspection sector.
What has prevented the take-up of vision for assembly guidance? Well, one industry where the take-up has been almost universal is in the assembly of printed circuit boards and in the manufacture of the integrated circuit packages which go on them. There, vision is "built in" to the machines, which carry out the assembly operations, and no one thinks of the vision systems as separate entities, they are just part of the overall functionality of the whole machine.
That still leaves a huge area of mechanical assembly where vision is not being exploited to anything like the extent forecast a decade or so ago. In those days, vision was:
(2) insufficiently robust; and
(3) too sensitive to changes in the component being handled, changes which were not material for manufacturing purposes but which changed the appearance sufficiently to upset the vision algorithms.
Lack of robustness showed itself partly in the purely physical sense lenses get knocked, electronics need a controlled environment and partly, in vision terms, with changes in ambient lighting affecting operations severely.
It was about ten years ago that a combination of faster processors and cleverer implementations of vision techniques, which until then had been known but were far too slow for industrial applications, began to offer ways of interpreting images which were completely unaffected by the absolute lighting levels and contrast in the scene; systems using these "normalised" techniques will continue to work when the light levels are too low to read a newspaper. Because these techniques do not need to look for edges or areas of a specific brightness in an image, they can cope with bigger changes in the components without being upset.
More recently, cameras which can cope with very high dynamic ranges of lighting for instance reading the printing on a light bulb envelope in the presence of the lit filament have been developed and are now commercially available at reasonable prices, and more conventional cameras have been improved to cope with varying light levels more effectively.
Costs have tumbled in vision systems in the last ten years, mainly because for most applications it is no longer necessary to design and manufacture (in relatively tiny quantities) specialised image-processing architectures. A particularly important breakthrough came with the advent of the PCI bus for the personal computer suddenly there was a bus fast enough to be used for moving image data around, which was manufactured by the millions rather than by hundreds or even tens. Processor power also increased to the point that for most vision tasks it was no longer necessary to use dedicated processors in addition to the host processor all the work could be done by the main (now only) host processor.
It is now possible to buy a complete vision system (an "intelligent camera"), with, in some cases, the lens incorporated along with the camera and processor, all in an industrially sealed enclosure, for between £2,000 and £4,000, from a variety of suppliers, each with their own particular competence and characteristics. It must be time to reconsider the use of vision in assembly not just to guide robots, but to check that humans or robots have correctly carried out a task before "adding value" and maybe concealing the error!
There could be another reason why take-up of vision for assembly automation has lagged behind expectations, and this is a laudable one, if true. In the mid-1980s we heard a lot of exhortations to "design for assembly", with examples such as the Japanese video recorders designed for assembly by selective compliance (SCARA) robots, well able to do their jobs without vision because the design made assembly simple. Has manufacturing industry learned enough in the last decade that no element of vision human or otherwise is needed to assemble any of today's products?
Some suppliers of "intelligent cameras" include:
Quiss, Germany: Tel: +49 89 800 1111; Fax: +49 89 800 2996; e-mail: firstname.lastname@example.org
RMV, Germany. Tel: +49 721 5603 0; Fax: +49 721 5603 33.
Siemens, Germany. Tel: +49 721 595 4076; Fax: +49 721 595 3997.
TIS, Germany. Tel: +49 361 230 10 0; Fax: +49 361 230 10 90; e-mail: email@example.com