Combination method of accelerated metric data search in image classification problems

The subject of research the paper is methods image classification on a set key point descriptors in computer vision systems. goal to improve performance structural by introducing indexed hash structures dataset reference images and consistent chain combination several stages data analysis process. Applied methods: BRISK detector descriptors, hashing tools, search large arrays, metric models for vector relevance estimation, software modeling. obtained results: developed an effective method based introduction high-speed using structures, that speeds up calculation dozens times; gain computing time increases with increase number descriptions; peculiarity classifier not exact performed, but taking into account permissible deviation from reference; experimentally verified effectiveness classification, which indicates efficiency proposed method. practical significance work construction transformed space representation, confirmation classifiers modifications examples, development applied implementing