VISION 2008 to showcase advances in multicore processors

Machine-vision software tools are greatly speeded up through the use of multicore-processor technology.

By Silvia Stoll, Messe Stuttgart,

Machine-vision software tools are greatly speeded up through the use of multicore-processor technology.

JUNE 19, 2008--VISION--The International Trade Fair for Machine Vision and Identification Technologies has established itself internationally within the last 20 years as the leading trade fair for machine-vision and identification technologies and will take place for the second time at the Stuttgart Trade Fair Centre (4 to 6 November; Stuttgart, Germany). Trade-fair visitors will benefit from the immense variety of products, systems, and user solutions for machine vision, all under one roof. The optimal solution may be found for the one or two user problems that initially appear to be unsolvable.

The development surge in previous years for multicore processors for computer systems has caused a stir among the manufacturers of software tools for machine vision. Multiprocessor technology is making it possible for users of machine-vision software to speed up analytical algorithms and, thus, achieve higher cycle speeds in the quality control of production processes.

"PC-based machine-vision solutions particularly benefit from multiprocessor technology," emphasizes Christian Demant, managing director of NeuroCheck, "as the availability of dual- and quad-core processors in this area is already very high. Something that cannot be forgotten," continues the software manufacturer, "is that the machine-vision software must first be installed using an intelligently implemented multithreading architecture for the advantages of the new processor technology to be really used"--exactly what has been put into practice by NeuroCheck. The new version, NeuroCheck 6.0, which will be presented at VISION 2008, offers full support for the optimized utilization of multicore hardware resources.

At last year's VISION trade fair, new possibilities for parallel processing of images by means of multicore processors were presented. MVTec Software had significantly optimized the automatic parallel processing in its standard software library HALCON so that the processing power of multicore processors can be utilized more efficiently and without additional effort for the software engineer. HALCON 8.0 automatically splits the data into the number of existing cores, processes the data separately, and joins the information together again.

"The additional speeding up is already implemented, for example, in high-speed applications such as bottle inspections, tool inspections, or in the semiconductor industry," says Lutz Kreutzer, manager for PR & marketing at MVTec. Actual multithreading for the processing at several processors requires software engineers with expert knowledge. "If, however, automatic parallel processing is performed on multicore processors, just as we have done," claims Kreutzer, "application and programming are significantly easier, and there is also the highest degree of reliability." The automation with HALCON shows comparable processing power as with the classic multithreading. However, according to Kreutzer, a lot of time and effort is saved as a result.

At the trade fair there will be software manufacturers such as STEMMER Imaging, which classifies the development surge with multiprocessors somewhat differently: "Multithreading and multiprocessor systems have already supported our software library, Common Vision Blox, since the first version to effectively resolve processing-intensive applications," says Peter Keppler, system solutions sales manager with STEMMER. "Of course, modern software must also support the new processor and computer technology," he continues, "yet even the highest processing power by far in a standard PC system is not offered by CPU, but rather by a GPU-graphic processor. Therefore, we have even ported multithreading onto the graphic card. Thus," he continues, "the previously almost inactive GPU processing power can be neatly used for the machine vision."Common Vision Blox is also prepared for tasks with the highest of processing power requirements," he states.

In addition to acquisition components such as cameras, frame grabbers, illumination, and lenses, machine-vision software plays a central role in solving an optical control task. It must keep pace with the production cycle and within a few milliseconds deal with many intelligent tasks simultaneously, for instance, bottle inspection filling level, the inscription on the label, and damage to the bottle or the shell. In doing so it must, for example, format the rough image data and help correct possible mistakes that arise through optics or nonoptimal illumination. It must be able to distinguish faults from natural structures, something that is often not easy to do. It must reduce the image data and extract and analyze the information that is relevant for the measuring task. Finally, it must be able to come to the correct decision: the object is OK or not OK. And it must send a signal on time to a programmable logic controller or a robot and separate the undesirable part, for example. Logging, documenting and archiving also belong to the function requirements of a machine-vision tool.

"Today computers are indeed faster, however, the data rates of current cameras and compilation technologies have also risen. Thus, special value must continue to be placed on an enduring variety of algorithms. There is still unbelievable potential here for speeding up," states Keppler. The area of 3-D machine vision is, for instance, a wide field and still offers immense possibilities for the future, to develop practical algorithms. "Therefore, STEMMER Imaging," continues the sales manager, "will be strengthened in the future and at VISION 2008 will offer solutions in hardware and software together with its partners."

MVTec also believes that the future is 3-D machine vision. "Our 3-D vision functions such as 3-D matching (go together) with only one camera, 3-D stereo, 3-D camera calibration, and 3-D alignment (coordination) are just as interesting for the robotics market. Of course, we will also further develop and create new algorithms for all the 2-D and 1-D technologies already established," says Kreutzer.

The importance of algorithms for NeuroCheck, can be seen, among other things, from the fact that the new software version NeuroCheck 6.0, which will be presented by the company at VISION 2008, contains fundamentally revised machine-vision algorithms to be able to optimally utilize the advantages of the multicore-processor technology. Demant has high expectations for the new multicore technology: "There is the possibility that new fields of application for machine vision will be developed."

In addition to speeding up computer processes and increasing of cycle times, there are, of course, many other requirements that the future-oriented software tools for machine vision must fulfill: for instance, hardware independence so that cameras from different manufacturers can be interchanged without any great expenditure. The flexibility of the software tools with regard to the resetting of measurement parameters and test values or the adding on of new features plays a large part, also the simple and intuitive application, as well as the guarantee of indirect compatibility and much more.

At VISION 2008, there will once again be "Industrial Vision Days," organized by the VDMA Specialist Department of Machine Vision, where proposals and answers will be given and solutions will be found. The expert forum is placed directly in the middle of all the trade-fair activity, so that visitors and exhibitors can benefit from the top-class presentations and interesting discussions.

More in Boards & Software