The machine vision industry is progressing at a rapid rate. The following sections provide an overview of where the technology
is today, and what we can expect in the near future.
Perhaps the largest trend in the industry is toward host-based
processing on PCs running in a Windows environment. The ever-increasing power of standard microcomputers, combined with open
architecture and standard development environments are proving sufficient for a large number of applications. This is resulting
in a decline in applications using specialized, embedded vision processors.
On the camera front, analog RS-170 cameras are still
the norm, but the use of digital, progressive-scan cameras is on the rise. High-resolution (megapixel) cameras are becoming
more common as many frame grabbers and vision boards now support them. Other trends include capturing only a portion of the
image from the detector to increase throughput, and the use of 10 and 12 bits of pixel data. Also expected to see increasing
use is newer serial communications such as Firewire (IEEE-1394), capable of delivering real-time (30 Hz) full-frame image rates and
As for lighting, traditional sources such as fluorescents and xenon strobes are still commonplace, but LEDs are becoming
the favored illumination source due to their long life and relatively low cost.
Communications options have also expanded for
machine vision systems, which are often a part of a larger automation system. Some systems now support Ethernet, DeviceNet,
and other factory-wide communication protocols.
Vision applications can now be developed faster than ever, thanks
to open systems and industry-standard development environments such as Microsoft Visual C++ and Visual Basic. Complete vision
libraries, available in .DLL and .OCX (ActiveX) formats, allow complex vision programs to be built into easy-to-use Windows applications.
Reusable vision code, combined with rapid application development environments, permit fast prototyping and evaluation of new systems
as well as substantially shortened development time.
New vision algorithms and improvements upon existing ones are also beginning
to appear. One example is advanced pattern recognition for object identification and location, which is now far more accurate
and robust than prior generations. Some new tools can now perform pattern finding on objects which vary arbitrarily in size
and rotation compared to the trained pattern, and which have substantial amounts of image degradation. These new geometric-based
pattern matching engines rely on object contours rather than object grayscale patterns. Another example is the increasing use
of color image analysis to perform classification and inspection tasks that were previously impossible or very difficult using traditional
grayscale methods. CONTINUE>>