Discussing problems of analyzing the properties of clustering methods to set specific image features in recognizing visual objects in computer vision systems. Was studied in comparative terms and methods of grouping the difference to the averagedefined quantization errors and other parameters, which allowed to establish and assess the possibility of their application. For specific image shows the results of experiments on the estimation accuracy of methods for different numbers of clusters.
computer vision, structural methods of recognition, key characteristics of the images, clustering description, differential clustering method, k-average method
"Yzuchenye svoistv metodov klasteryzatsyy prymenytelno k mnozhestvam kharakternыkh pryznakov yzobrazhenyi" [Study properties of clustering methods relative to the set of characteristic signs of images],
Information Processing Systems,