SELF-LEARNING METHODS IN THE SPACE OF IMAGE STRUCTURAL SIGN CRITERIONS

摘要

The issues of enhancing efficiency of the structural image recognition methods in the computer vision systems are discussed. For systematization of the space of signs it is suggested to perform self-learning with application of differential grouping methods and Kohonen networks. As the result, it is developed a more efficient in terms of processing fast-action vector description of the standard bases. The computer modeling results are provided for estimation of the quality of learning and the methods operation capabilities for the application image database.