Optimization of Identification of Micro-Objects by Correlation Functions of a Random Sequence of Image Points

A methodology has been developed for the identification, recognition and classification of micro-objects based on the use of tools for extracting statistical, dynamic, morphometric, correlation and other specific characteristics of images, such as pollen grains, unicellular microorganisms, medical objects, images of useful minerals and others. The mechanisms for identifying micro-objects are proposed based on tools for selecting characteristic parts and points of an image, fixing reference points, linking points of differences, transitions, adaptive segmentation and extracting features of constituent components in the image structure. The tools for filtering images of micro-objects under internal and external influences, in the presence of noise (interference), blur, point shifts and other defects were investigated. The tools for tracking, detecting and correcting defective points are proposed, including modules for reducing zero points, regulating the start, center, segment boundaries, the width of the reference point mask, and the color-brightness picture of the image under conditions of low robustness based on stochastic models of image identification. Highly accurate identification mechanisms are obtained, which are performed at low cost using a variety of correlation functions and non-stationary properties. A set of programs for identification, recognition and classification of images of micro-objects has been implemented in the C++ language in the CUDA environment.

Hali tarjima qilinmagan