Machine vision

Journal Article · Thu Jun 01 04:00:00 EDT 1989 · Mech. Eng.; (United States)

OSTI ID:5909643

Horn, D

To keep up with the speeds of modern production lines, most machine vision applications require very powerful computers (often parallel-processing machines), which process millions of points of data in real time. The human brain performs approximately 100 billion logical floating-point operations each second. That is 400 times the speed of a Cray-1 supercomputer. The right software must be developed for parallel-processing computers. The NSF has awarded Rensselaer Polytechnic Institute (Troy, N.Y.) a $2 million grant for parallel- and image-processing software research. Over the last 15 years, Rensselaer has been conducting image-processing research, including work with high-definition TV (HDTV) and image coding and understanding. A similar NSF grant has been awarded to Michigan State University (East Lansing, Mich.) Neural networks are supposed to emulate human learning patterns. These networks and their hardware implementations (neurocomputers) show a great deal of promise for machine vision systems because they allow the systems to understand the use sensory data input more effectively. Neurocomputers excel at pattern-recognition tasks when input data are fuzzy or the vision algorithm is not optimal and is difficult to ascertain.

• OSTI ID: 5909643
• Journal Information: Mech. Eng.; (United States), Journal Name: Mech. Eng.; (United States) Vol. 111:6; ISSN MEENA
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210* -- Supercomputers-- (1987-1989)
ALGORITHMS
BODY
BRAIN
CENTRAL NERVOUS SYSTEM
COMPUTER CODES
DATA ACQUISITION SYSTEMS
IMAGE PROCESSING
LEARNING
MAN-MACHINE SYSTEMS
MATHEMATICAL LOGIC
MEDICINE
NERVOUS SYSTEM
NEUROLOGY
ORGANS
PARALLEL PROCESSING
PATTERN RECOGNITION
PROCESSING
PROGRAMMING
REAL TIME SYSTEMS
VISION