Optimization of Recognition of Microorganisms Based on Histological Information Structures of Images

Constructive approaches, principles, methods of identification, recognition, and classification of micro-objects with the isolation of histological microparticles and the use of geometric characteristics of images have been developed. The technology of preliminary processing of images is investigated and modified on the basis of mechanisms for extracting statistical, dynamic, specific characteristics of information, setting the values of variables, checking the similarity of micro-objects. Mechanisms for identifying contours, dividing them into segments (classes), using histological redundant information structures, recognition and classification based on a set of numerical characteristics, a database, an image database, a knowledge base, and rules are proposed. The algorithms of learning three-layer neural network, Kohonen neural network, radial-basis network, which is combined with orthogonal algebraic polynomials 5 and 7, the interpolation spline function of Daubechies 7, mechanisms for setting variables along the boundaries of acceptable values, are studied. Mechanisms have been developed for classifying images of microparticles on the basis of histological redundant information structures of images of medical objects, segmentation, highlighting informative points, comparing redundant information structures of the original image and a modal example, and searching for characteristic features. The efficiency of the generalized algorithm for the identification of micro-objects is investigated according to the criteria of information processing performance and the correctness of recognition of micro-objects. A software complex for identification, recognition, classification of images has been developed, which has been tested taking into account the conditions for processing information from medical objects presented for diagnosing and predicting tuberculosis diseases.