METRIC ALGORITHMS OF SYMBOL DETECTION FOR DETERMINING A GROUP OF SIGNIFICANT OBJECTS

This article examines the algorithm of nearest neighbors, which widely uses metric classification algorithms and the combinatorial theory of reconnection with the precedent information in solving practical recognition problems. One of the most important issues in the development of image recognition systems is that the image recognition system determines the characteristics of objects. Most pattern recognition methods are used in the process of recognizing not all objects of the training sample, but a certain part of the objects. The formation of a training sample by filtering objects according to the level of importance for metric classification algorithms is carried out by summing combinatorial estimates with the possibility of analyzing the compactness profile when choosing basic objects. One way to modify the nearest neighbor algorithm is to select a subset of a small number of base objects and recognize control objects using this subset of base objects. A criterion for the effectiveness of the use of combinatorics formulas in the implementation of adaptive control and the concept of compactness of the selection profile are proposed. The proposed algorithm is on the opposite side compared to the algorithms of the STOLP category.

Перевод пока недоступен