Procedure of classification of patterns in the scheme of confluent analysis with correlation of errors in results of estimation of patterns coordinates

Abstract

In the paper the problem of construction of dividing line for two classes of the images set by two-dimensional signs is considered and solved. The case when all signs are burdened by random and regular errors that leads to a specific correlation picture of experiment, set in advance by means of a matrix of the correlation moments is thus considered. By means of numerical simulation the accuracy received in work on the basis of modified method of the maximum credibility of estimations of parametres of a dividing line is shown.