Vision systems improve performance at less cost
George Kotelly
Executive Editor
Despite the numerous performance advantages of machine-vision systems, developers are always tempering technology achievement with the requirements of affordability and availability. To spur the continuous growth of vision-system sales, developers are constantly seeking to reduce costs, make products smaller, improve quality, and shorten time to market. These guidelines are being applied across the vision-application spectrum, such as grading pharmaceutical liquids, placing parts with robots, or packing functions into smaller boards.
In the pharmaceutical industry, machine-vision product-grading systems ensure that contaminants have not infected medical vials. To achieve ever-higher medical-product quality at reasonable cost, a vision inspection module has been built that incorporates robust microprocessor technology with frame-grabber operation limited to image acquisition. According to contributing editor Winn Hardin, this module has been integrated into a PC-based system running under Windows NT to achieve fast imaging defect analysis of pharmaceutical liquids.
To gain production savings, semiconductor manufacturers have been longtime users of vision-guided robotic systems to ensure correct parts placement on printed-circuit boards. As spotlighted by editor at large Andy Wilson, these suppliers are now eliminating costly part fixtures by integrating third-party machine-vision tools into their robotic simulation software. With this approach, they can simulate robotic and image-inspection tasks off-line and save development costs.
With the increasing density of integrated circuits and the availability of standard PC-based buses, image-processing board suppliers are incorporating increased functionality into single boards. In this month`s Product Focus, editor at large Andy Wilson discusses how compact PC-104 and PC-104+ boards are now capable of providing numerous functions such as networking, analog and digital conversion and compression, monitor display, and motion-control. What`s more, these self-stacking boards allow easy expansion without backplanes and card cages.
In this month`s Software Vision column, Mark A. Vogel, president of Recognition Science Inc., expands on last month`s coverage of vision-based pattern-recognition techniques by discussing the discriminant information needed to rank-order pattern-recognition features and build classifiers using subsets of the highly ranked features. This can help to minimize development time and to define an optimum classifier.
