Home – Home – Jornals – Avtometriya 2006 number 6

The problem of image reconstruction is considered for the case when the recording system point-spread function and the recorded image are known with random errors. An efficient computational image reconstruction algorithm that uses a 2D Discrete Fourier Transform and a Fast Fourier Transform algorithm is designed. An algorithm for reliable estimation of the optimal regularization parameter is proposed to choose the regularization parameter determining the reconstruction problem solution accuracy. The reconstruction algorithm can be adapted to identify a point-spread function of dynamic image recording systems.

Algorithms for time-finite signal restoration are developed in view of feasibility of the pulse characteristic of processing devices. Functions with a lower level of side lobes than in the known functions for single-channel and two-channel restoration systems with using signal samples and time-finite signal derivatives are synthesized. Optimum functions are calculated and quasi-optimum time-finite synthesizing functions of the interpolation algorithm are proposed, which allow consideration of time constraints of initial signal realization and feasibility of the processing devices. It is found that the proposed algorithms will reduce the root-mean-square error of interpolation by 45

A method is proposed to construct nonlinear nonparametric pattern recognition algorithms for efficient solution of classification problems in the conditions of small samples. Their properties are analyzed using results of computational experiments.

A system of likelihood equations for optimal estimation of optical pulse parameters in view of shot noise is obtained. Algorithms for estimates of the amplitude and duration of a Gaussian pulse, and also its position on the time axis are found. The Cramer-Rao bounds for non-joint and joint estimates are determined.

The main results of comprehensive analysis of possible ways for solving the problem of forecasting the annual inflow of the River Ob at the site of the Novosibirsk Hydroelectric Station are considered. The research is based on ideas and methods of variant modeling. A set of probabilistic models is constructed for inflow forecasting with different accuracy and term (the period between the date of forecast and the beginning of the forecast period). Ways for future research are outlined.

Projective mapping of physical space on the spheroid domain, that is, a hyperbolic Hilbert space model, is considered. Straight lines, diameters, perpendiculars, distances, and angles on the Hilbert plane are determined. An image of visual space inside the spheroid is shown as well as its conformal mapping on the Poincare model. A formula for apparent size of the Moon elevated over the horizon is presented.

A computer system modeling the effectiveness function by results of independent tests based on the nonparametric kernel regression estimate method is developed.

The influence of light divergence and the shape of light scatters of an object with volume scattering on its brightness is analyzed.

Numerical formulation of the incident wave source conditions for a domain enclosed in a nonuniform envelope within the scope of the TF/SF technique (Total-Field/Scattering- Field technique) is proposed. This approach is compared with a traditional analytical approach to defining the incident wave source conditions. Benefits of the approach are shown.

Temperature dependences of the Hall coefficient and magnetoresistance in heteroepitaxial structures of p-Cd_xHg_x-1Te (MCT) of X_CdTe = 0.22

Results of numerical modeling a new type of acousto-optic (AO) filters employing multibeam expanders are discussed. A planar filter design in thin m) optical waveguides of chalcogenide glass on lithium niobate is proposed. Analysis of the devices is carried out in a spectral approach with application of numerical experiments by the finite-difference time domain (FDTD) method. It is shown that the proposed multi-reflector optical elements possess high dispersion properties such that the AO filter with a small size (below 0.1 cm) ensures a record thin linewidth (~1 nm) at a wavelength of 1.54 m, a tuning range greater than 40 nm, and a switching time of less than 2 s. In addition, the normalized linewidth per unit length of the device is 20 times better than the theoretical limit for available types of AO filters developed on the same substrate and by a common technology but without multibeam expanders.

An improved optoelectronic small linear displacement sensor based on laser beam diffraction on two phase diffraction gratings is proposed and investigated. The possibility of doubling the region of linearity while detecting signals in the first two diffraction orders is shown. A device for measuring surface irregularities is designed using this sensor. Experiments on measuring surface profiles of relief structures are carried out.

A stabilized fiber optic sensor for remote measuring the angle of object inclination from the vertical has been developed. The measurement principle is based on beam refraction by a transparent free hanging plane-parallel plate as the angular orientation of the sensitive element case changes. Output signal stabilization is assured by using two spatially separated measuring channels.