Accession Number : AD0767273

Title :   The Mucciardi-Gose Clustering Algorithm and Its Applications in Automatic Pattern Recognition.

Descriptive Note : Final rept. 15 Mar-15 Nov 72,

Corporate Author : ADAPTRONICS INC MCLEAN VA

Personal Author(s) : Mucciardi,Anthony N.

Report Date : JUL 1973

Pagination or Media Count : 75

Abstract : A procedure known as the Mucciardi-Gose clustering algorithm, CLUSTR, for determining the geometrical or statistical relationships among groups of N-dimensional vectors is presented. The vectors may be thought of as samples from some complex process that is under study. For example, the process may be aerial reconnaissance photography, and the vectors may be digital representations of the pictures. In this example, the geometrical or statistical relationships between the pictures, some part of the pictures, or some derivative of the pictures, must be known before an automatic analysis of the content of the pictures can be performed by machine. The CLUSTR algorithm provides a means of determining these relationships. A general discussion of clustering algorithms is given; the particular advantages of the Mucciardi-Gose procedure are described. The mathematical basis for, and the program structure of, the CLUSTR algorithm are presented in detail. Topics covered include: initial cluster hypervolume estimates, birth and growth rate processes of clusters, estimation of cluster hypervolume overlap. Specific instructions for use of the programs and for interpretation of the results are provided. (Author)

Descriptors :   (*PATTERN RECOGNITION, ALGORITHMS), MULTIVARIATE ANALYSIS, PROBABILITY DENSITY FUNCTIONS, COMPUTER PROGRAMMING, AERIAL RECONNAISSANCE, PICTURES, MATHEMATICAL MODELS

Subject Categories : Computer Programming and Software

Distribution Statement : APPROVED FOR PUBLIC RELEASE