Optimization of Micro-Object Identification Based on a Pyramidal Model with Image Segmentation

A methodology has been developed for optimizing the processes of identification, recognition and classification of micro-objects, pollen grains, unicellular microorganisms, and minerals in rock mass based on the use of properties and characteristics of components in the image structure. Tools for reducing the error in identifying micro-objects are proposed based on filtering defective points using statistical, dynamic, morphometric, geometric, histological, spectral-frequency, and brightness characteristics of the image. The efficiency of the mechanism with the execution of the operation of tracking, detection, correction of defective points, detection of growth dynamics, visual delineation, dynamic approximation, smoothing based on the use of Gaussian, median filtering, Fourier, shift - transformations is investigated. The processes of identifying micro-objects have been optimized based on the alignment of histology sections, finding fragment contours, a set of levels and thresholds, color coding and visualization of micro-objects. A software package for identification, recognition and classification of micro-objects was implemented and software modules were tested on real measurements of cellular elements of the inflammatory series of lung diseases.