In article the short description of the most often used methods of classification at pattern recognition is given. The main attention is paid to the methods allowing development of a system for image recognition in a real time scale. The features formation method on the base of two-dimensional spatial spectrums of objects images is offered and application of similarity metrics in a decision-making rule for image classification is described.
Experimental data of correct and erroneous recognition probabilities as well as image classification time depending on a number of features and on the identification threshold value are presented and analyzed