DEVELOPMENT AND METHOD OF SIMULATION OF CERTAIN ALGORITHMS OF NORMALIZATION OF PLANE PATTERNS ON A COMPUTER (ETsVM),
Abstract
As the efficiency of pattern recognition is greatly enhanced by preliminary normalization (orientation, size standardization, and centering) of the test pattern, the article examines normalization procedures and algorithms with a special emphasis on orientation. Normalization of input situations eliminates a part of noise from the signal parameters. Mathematically, the pattern normalization is the finding of such an algorithm which minimizes the dispersion of parameters of realizations in the space on indicants. Orientation of a rectangle and a triangle are considered; the involved algorithm is invariant with respect to pattern size and translation. An algorithm of plane-pattern orientation based on a minimum-area circumscribed rectangle was simulated on an Ural-2 digital computer; numerals 2 and 3 written by hand in various positions on a 39 x 39-element square were tested. Operators for pattern size standardization and centering are briefly discussed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 02, 1970
- Accession Number
- AD0705846
Entities
People
- B. V. Anisimov
- V. K. Zlobin
Organizations
- National Air and Space Intelligence Center