Mind over matter
Andy Wilson Editor
Pixar Animation Studios is now famous due to the "overnight success" of John Lassiter`s Toy Story. But more than a decade ago, Pixar perfected its Image Computer--a machine specifically designed to render 3-D images. Although graphics algorithms ran faster on this machine, it would be Sun SPARCstations that rendered Toy Story more than a decade later. The hardware changed, but the algorithms remained the same.
However, many developers of vision systems do not have a $30 million budget or two-year development cycles. As John Haystead points out in this month`s Product Focus (p. 40), VME image-processing boards are still required to handle large amounts of data in close to real time. Because of this, such systems include multiple processors, large memories, and high-speed I/O. Integrating these subsystems is not easy.
In developing a VME-based system to inspect meat (p. 26), Giuseppe Longobardi of the Centro di Eccellenza Optronica in Italy fused multiple images from visible, infra-red, and x-ray detectors and sensors. Using many off-the-shelf components, the system integrates detectors, frame grabbers, and a custom-built DSP-based image processor. Off-the-shelf CPUs were also ruled out by developers at GDE systems (San Diego, CA), where multiple image and array processors generate digital maps for cruise missiles (p. 20).
Hardware-assisted image processing is also providing windowing/leveling, rotation, and look-up functions, as Rick Nelson explains in his article on display controllers (p. 30). In less-demanding applications such as microscopy, general-purpose CPUs are coming to the forefront. Macintosh and PCs are hosting frame grabbers and analog-to-digital converter boards in SEM, TEM, and optical microscopes (p. 16).
But while most embedded image-processing systems still require integration, it is software that drives new applications. In our Spotlight on Advanced Technology, Barry Phillips discusses how fuzzy-logic algorithms are helping researchers develop novel medical and weather-forecasting applications (p. 34). As general-purpose processors increase speed and I/O capability, vision systems will be deployed more frequently. But it will be the algorithms and the people who crafted them that will be remembered, not the computer architectures they were built around.
