THE DATA CONSOLIDATION AND DECISION MAKING BY THE RANKING OF GROUPS WITH IDENTICAL CLASSIFICATION IN HYBRID RECOGNITION SYSTEMS
Hybrid recognition systems based on the use of a set of patterns for single recognition object.
Its patterns have a different origin and allow to obtain solutions in a wide range of changes in external conditions, noise level and distortions of the input data also their heterogeneity. High recognition result reliability is ensured by the presence of an extended set of signs by which a decision is made. However, in this case there is also a negative aspect - an increase the time complexity of analysis and decision making. Wherein, various signs may have different informational content for the formation of a reliable decision on the classification depending on the current conditions of observation of the object of recognition. Researches of the subject area showed that to ensure high reliability of the classification with minimal time spent, it is necessary to consolidate the data. The essence of consolidation, in this case, is to make a decision based on an analysis not of the entire set of data, but only of the set of the most informative ones. In this case, the information content of the same data may vary depending on the external conditions for obtaining data. Thus, the aim of the researching is developing a new method of heterogeneity data consolidation and decision making in hybrid combined recognition systems that allows to exclude less informative data from the analysis. It leads to reducing the time complexity of the decision-making process. The developed method is based on the formation and ranking of groups of patterns with identical classification, which allows to select the most informative of them for analysis and decision making. According to the developed method, classification is carried out according to different patterns of the recognition object and groups of patterns are formed that give a similar result. Obviously, in the ideal case, all images should give the same classification result, but in practice this is not always the case. Further, groups of images are ranked in quantitative terms and the most numerous group arrow to the most reliable result. Thus, the decision is made on the basis of the classification results provided, from the most reliable option (the most numerous group) to the less reliable.