Smelyakov K. Models and methods of segmentation of irregular object images for off-line machine vision systems

The thesis is devoted to solving a scientific problem of ensuring the required level of stability in segmentation of textured and homogeneous irregular object images, that are obtained under the stipulation of significant variations of their topological, geometrical and photometric parameters, on the ground of development of a system of mathematical models and computing methods that provide effectual recognition of the segmented images of irregular objects in off-line machine vision systems operating in the conditions of uncertainty. To solve this problem it is developed a structural scheme of adaptive segmentation for off-line machine vision systems, together with a system of efficient mathematical models, criteria and computing methods which provide adaptive segmentation of textured and homogeneous images of irregular objects that are obtained in the conditions of uncertainty.