ADAPTIVE FILTERING AND GRADIENT OPTIMIZATION MECHANISMS FOR ACCURATE IDENTIFICATION OF MICRO-OBJECT IMAGES

I. Jumanov, R. Safarov, O. Djumanov

Samarqand davlat universiteti ilmiy tadqiqotlar axborotnomasi · 2026-yil

Annotatsiya

A methodology has been developed for optimizing the identification, recognition, and classification of micro-objects using redundant image information structures - histological, morphological, fractal, correlation, and spectral characteristics. The following mechanisms are implemented: transformations, scaling, reconstruction of images of selection and segmentation of the contour; selection of control points on the contour, setting the values of variables, taking into account their statistical relationships, the dynamics of change; shift, rotation, deformation of a sequence of segments. Researched neural network architectures: 19-13-12; 30-40-12; 40-15-12; 35-15-12; 25-12-12 with learning algorithms for the neural network based on the Levenberg-Marquardt method. A software package for visualization, recognition, and classification of images of micro-objects is implemented, based on the use of mechanisms for extracting correlation and spectral characteristics, a model for adapting correlation windows, optimization, and modified training of a multilayer neural network.

Maqola ma’lumotlari
MualliflarI. Jumanov, R. Safarov, O. Djumanov
JurnalSamarqand davlat universiteti ilmiy tadqiqotlar axborotnomasi
Nashr sanasi2026-07-01
Jild1
Son3
Betlar102-110
DOI10.59251/2181-3973.2025.v1.138.1.3950

Samarqand davlat universiteti ilmiy tadqiqotlar axborotnomasi jurnalidan boshqa maqolalar

Samarqand davlat universiteti ilmiy tadqiqotlar axborotnomasi — barcha maqolalar