Abstract shape recognition by machine

  • Authors:
  • Mary Elizabeth Stevens

  • Affiliations:
  • National Bureau of Standards, Washington, D.C.

  • Venue:
  • AFIPS '61 (Eastern) Proceedings of the December 12-14, 1961, eastern joint computer conference: computers - key to total systems control
  • Year:
  • 1961

Quantified Score

Hi-index 0.00

Visualization

Abstract

Graphic pattern recognition in the development of total systems for information selection and retrieval is considered. A particular hypothetical machine model for shape recognition is described. Examples are given of recognition of 12 to 20 categories of geometric shapes, by use of a contour-projection principle. Within certain limitations, the model identifies these graphic patterns, regardless of size, location, and certain rotational transformations. It provides means for detecting classes of patterns which would or would not be ambiguous with respect to any of the patterns that are recognizable. Problems in practical application, including possibilities for recognition of constrained handdrawn figures, are considered. Possible means of implementation are also discussed.