Peripherals push performance of vision systems
Andy Wilson, Editor
Many advanced electronic imaging systems need to digitize, process, transfer, and store vast quantities of digital images. Because of this, OEM subsystems used in vision systems often consist of state-of-the-art digitizers, parallel processors, networking and RAID or optical subsystems. In manufacturing, for example, processes may involve aligning individual objects in 3-D space. To do so may require the use of laser-based scanning systems, as Kevin Harding of the Industrial Technology Institute (Ann Arbor, MI) explains (see p. 32).
Sometimes processing such data is the role of the pipelined or distributed processor. Because of this, board-level vendors are offering attached coprocessors, parallel processors, and DSP-based designs on buses for PCI, CompactPCI, and VME buses. Contributing editor Rick Nelson reports that many of these attached coprocessor boards and modules incorporate multiple processors interconnected via high-throughput data paths (see p. 56).
Some medical image-processing systems use coprocessors to visualize two-dimensional (2-D) data in three dimensions. This requires the development of sophisticated reconstruction and interpolation algorithms, coupled with an inherent knowledge of sensor dynamics. As Larry Brown reports, computer-aided tomography is just one modality using high-performance image processing to reconstruct 3-D data (see p. 44).
The military is also using high-performance computers to analyze electromagnetic and thermal signatures from targets to help weapons-systems designers create smarter systems. Contributing editor John Haystead reports that the USAF is currently using high-performance workstations to perform such tasks at Department of Defense test ranges (see p. 38).
High-performance workstations are also being used at the crystallography facility of Lawrence Livermore National Laboratory, where researchers are analyzing molecules to determine the crystal structures of proteins. Contributing editor John Haystead discusses the design of a networked CCD-based x-ray crystallography system (see p. 28).
Developers are always looking for novel image-compression techniques to minimize image data and reduce storage system costs. Contributing editor Barry Phillips examines the role of image-compression standards and takes a look at the latest in fractal- and wavelet-based algorithms (see p. 50).
