Home – Home – Jornals – Avtometriya 2004 number 4

Large-dimension signals and images are frequently reconstructed by means of iterative algorithms in which the regularization parameter is the number of iteration at which the construction of the minimizing sequence stops. The choice of the stopping iteration is an important problem. Its solution requires substantial a priori information unavailable in practice. The present paper, using the cross-validation method, proposes an algorithm for estimating the number of iteration at which the solution has the minimum error.

The maximum likelihood method of simultaneous detection of an inhomogeneous image and estimation of its area in the presence of a space noise is considered. Exact characteristics of the algorithm performance are found. The obtained results are applied to an image in the shape of ellipse with a linearly varied intensity.

An algorithm for segmentation of anisotropic images and construction of a field of directions, which is based on analyzing integral characteristics of a local energy spectrum calculated in a sliding window, is proposed. The efficiency of applying the algorithm to processing images of a microrelief of traces of real objects identified while studying clues is illustrated.

A hydroacoustic signal processing method is considered. The method uses an adaptive environment model including the medium of propagation and the radiation sources contained in it, and also a receiving antenna array model. Results of investigation on a software test complex are presented. It is shown that the method can be successfully applied to problems of primary hydroacoustic signal processing.

Peculiarities of application of broadband emission detectors in spectral ratio pyrometers are considered. To calculate the temperature, it is proposed to use the Planck formula that ensures a high measurement accuracy over a wide range of temperatures.

A new type of integrated-optic polarization splitter is proposed. Its operation is based on significant mode birefringence of a four-layer Y-brancher with specially defined parameters. The artificial birefringence is formed by combining a weakly guided waveguide and an abruptly inhomogeneous waveguide. The main parameters of the device are estimated by the finite-difference beam propagation method.

A simplified method for evaluation of potential distribution in thin graded gap layers of a photosensitive semiconductor is proposed. Formation of space charge regions in a semiconductor film with a graded gap layer in its middle part, and also with graded gap layers near external surfaces, which are used to control the characteristics of planar photodiodes and to suppress the surface recombination in IR photodetector arrays based on epitaxial MCT layers is considered in a linear approximation.

The modes of pulse transmission in a fiber-optic communication line with in-line nonlinear devices, that is, nonlinear optical loop mirrors (NOLM), are considered. The ability of NOLM to regenerate the pulse shape and suppress spontaneous emission noise is investigated. The possibility of pulse stabilization over a wide range of parameters of the line (optical filter bandwidth) is shown.

The functioning of a distributed computing system in the mode of processing a packet of parallel tasks is considered. A modification of one sequential stochastic algorithm and its parallel implementation are proposed. Results of modeling are presented. They confirm that the obtained algorithms ensure at least a subminimum time of solving tasks

It is shown that results of the conventional fault-tolerant system modeling approach based on architecture analysis proceeding from reliability characteristics of elements do not characterize the fundamental properties of the architectures and are of special character. The proposed generalized statement of the problem of fault-tolerance analysis is based on the axiom of inevitability of faults of system elements (including independent multiple and

An algorithm based on evolutionary computation and templates of functions is described. It is intended for discovering analytical descriptions of new dense families of optimal regular networks. New families of regular parametrically described networks with degrees 3 and 6, which improve the known results, are presented.

A problem of creating a dynamic model for market trends, capable of rapid adaptation to various forms of external adjustment of the rate of securities, is posed and solved.