CRITERIA FOR THE DETERMINATION OF INFORMATIVITY AND PATTERNS RANKING FOR MAKING DECISIONS IN MULTI-PARAMETRIC COMBINED RECOGNITION SYSTEMS
One of the most promising concepts for construct recognition systems is combined systems of patterns recognition. Its concept proposed by Zhuravlev Yu.I. In this combined systems of patterns recog- nitions objects description included the signs of different types of information (deterministic, probabilistic, logical or structural). In this case, a sufficiently large-scale and universal database is created to describe any recognition objects for a variety of conditions for their observation. But this concept does not provide high level of recognition authenticity. For eliminate this disadvantage may be used multi-parameter com- bined recognition systems where the signs have a different nature of occurrence. However, such systems also have a significant disadvantage: essential increase the number of information signs for the recognition ob- jects. This article considered questions of the informative criteria and patterns ranking development for decision making in multi-parameter combined recognition systems. The proposed criteria allow at any time in
the system’s life cycle to carry out selection of informative patterns for high level of recognition authenticity
providing. The proposed approach of two-stage ranking of informative images for the object of recognition. It allows determine the further strategy of their analysis and decision-making. Determination informativity criteria is based on stationarity and non-stationarity distortions of object’s patterns when they are reflected
on sign’s area. It allows for any patterns of recognition objects to use two criteria (stationarity and non- stationarity) for determination of informativity. At the first stage carry out a priori determining patterns in- formativity with stationary criteria and its information is base for second stage – a posteriori determining informativity by non-stationary criteria. These two stages allow make the choice most informativity patterns for next decision-making. It allows reduce the information field of analysis and speed up result obtaining. Also on the two-stage informativity determining based patterns ranking, which was proposed. For decision- making speed maximization informative patterns analysis is carried out in accordance with ranking and de-
creasing of informativity level. Thus, at any time of system’s life cycle, the decision-making is based on the most informativity patterns of recognition object.