Abstract:  Recognizing the image shapes of the groups of point and / or small objects (TRP) is a non-trivial task due to the incoherence and degeneracy of their elements. The task becomes furthermore complicated for TRPs with non-stationary configuration, such as "chains" and "congestions". The differentiation of these types of images holds an independent value, and it can also be used  in order to branch out the algorithm for more detailed recognition of the TRP. In order to synthesize effective chain and clusters differentiators, it is important to determine the principle for describing the TRP form, as well as  discriminatory characteristics, to determine the statistics and characteristics of decision-making in the presence of disrupting factors. The solution of this problem is achieved by the methods found within the theory of processing digital images and signals, the theory of contour analysis for the synthesis of algorithms used in order to describе and analyzе the shape of images, methods of probability theory and mathematical statistics for the synthesis of decision-making methods. The procedure for constructing a minimal spanning tree is used in order to link isolated TRP elements to a single object. Its shape is described by a chain complex-valued code which is its contour. The dependence between the width of the energy spectrum of such a contour or the value of the correlation interval of its readings and the degree of complexity of the form allowed the authors to choose the distinction between chains and clusters of the characteristic of the autocorrelation function (ACF) of the contour as a discriminating feature . The width of the ACF (correlation interval) and the correlation of neighboring elements of the contours of the TRP are studied as such characteristics. The corresponding algorithms for distinguishing TRTs of these classes as chain identifiers are synthesized. The characteristics of decision algorithms for various observation conditions are found. A comparative analysis of their effectiveness and applicability limits is also performed.

Abstract: The issues arising in the implementation of the technical approach to the recognition of the group point object (GPO) images on the basis of the binding elements of the continuous associated images (CAI) are considered. The basic CAI formation models of the point scene defocusing are analyzed. For bell-shaped and rectangular impulse response obtained defocusing filter selection rules limiting circuit CAI level with maximum fault tolerance quantization of the image brightness. Concretize methodology harmonization CAI model parameters with the density of the GPO members. For clarity and simplify the operator interface work with models CAI highlight - radius of the filter impulse response the ratio of the radius with a threshold localize spatially compact objects is obtained. Synthesized numerical method for the formation of base procedure ASO, characterized by one - two orders of magnitude higher performance than with an approach based on the fast Fourier transform for compact GPO. The method is based on the filtering properties of the deltoid brightness distribution GPO elements and limitations of the low pass filter window size based on the number of quantization levels of its impulse response. The results are aimed at ensuring the quality of the detection procedures, permits and GPO recognition in noisy images, set in a one-dimensional, two-dimensional and three-dimensional spaces in their technical implementation.