Publishing House SB RAS:

Publishing House SB RAS:

Address of the Publishing House SB RAS:
Morskoy pr. 2, 630090 Novosibirsk, Russia



Advanced Search

Avtometriya

2006 year, number 1

1.
Adaptive filtering with subpixel estimating coordinates of point targets

V. S. Kirichuk, V. P. Kosykh, and T. Kurmanbek uulu
Novosibirsk
Pages: 3-11

Abstract >>
An algorithm for searching and subpixel estimating coordinates of point targets by images received from photodetector arrays is considered. Results of numerical modeling and an estimate of efficiency of the proposed algorithmic solutions are presented.



2.
On non-parametric methods of applied bispectral analysis

V. G. Alekseev
Zvenigorod, Moscow Region
Pages: 12-19

Abstract >>
The most interesting results and recommendations on the bispectral analysis of discrete- time stationary random processes are collected in this paper. Exact definitions of both the bispectral density and the related mathematical objects are presented.



3.
Adaptive anisotropic filter for determining elements of intersecting by a contour signal a zero level

A. M. Prudnikov, I. V. Razin, and V. S. Emdin
St. Petersburg
Pages: 20-26

Abstract >>
A two-dimensional adaptive filter that forms a contour signal of image objects by accumulation of the initial signal in the direction of orientation of the contour difference and by its double differentiation in the orthogonal direction is implemented. The adaptation parameters of the filter are local estimates obtained by procedures of for- mation of a covariance matrix, its diagonalization, and estimation of eigenvectors of the characteristic equation of the matrix determinant. The signal accumulation operation compensates for the loss in the signal-to-noise ratio of the filter contour signal, which is explained by the necessity of performing the double differentiation operation. This is a favorable advantage of the filter over the classical Marr-Hildrett filter.



4.
Development and analysis of hybrid algorithms in pattern recognition problems

A. V. Lapko, V. A. Lapko, and S. G. Yaroslavtsev
Krasnoyarsk
Pages: 27-33

Abstract >>
A technique is proposed for synthesis and analysis of hybrid models for effective usage of a priori information on the kind of equation of separating surface between classes and information of the training sample.



5.
Dependence of the embedded diagnostic structure robustness on the procedure of addressing the computing system modules

Yu.K.Dimitriev
Novosibirsk
Pages: 34-46

Abstract >>
Dependence of the structural robustness of the embedded diagnostic circulant graph on the vertex numbering of the host graph of computing system interconnections is studied. The host graph is a graph of direct product of cyclic subgroups. The embedding is based on partitioning a group into cosets via a cyclic subgroup of the chosen group element degrees. It is found that for the proposed methods of ordered numbering on a group graph, which are known as coordinate methods, by the evenness of cyclic subgroup orders and the evenness of the coset number, one can judge a comparative value of the used index of structural diagnostic graph robustness. Analysis of coordinate numberings is carried out to establish preference of certain numberings over others with respect to the structural robustness index.



6.
Formal logical reasoning approach and mamdani

I. A. Khodashinsky
Tomsk
Pages: 47-58

Abstract >>
Fussy estimation procedures based on the formal logical reasoning approach and Mamdani



7.
Algorithms for distributing task packets with variable parameters on computing system machines

M. S. Sedelnikov
Novosibirsk
Pages: 59-66

Abstract >>
Distributed computing systems functioning in regimes of processing a parallel task packet with variable parameters is considered. Heuristic algorithms for distributing task packets on elementary machines of a computing system, which minimize the time or penalty for solution delay, are proposed. Modeling results confirming that the algorithms ensure at least subminimal values of criterion functions are listed.



8.
Sequential classification procedure for processes of autoregression with random coefficients

D. V. Kashkovsky
Tomsk
Pages: 67-76

Abstract >>
For classifying stable processes of autoregression with random coefficients and Gaussian noise, it is proposed to use a sequential procedure with a guaranteed probability of right decision. The lower bound for the probability of right classification and an asymptotic formula for mean procedure duration are obtained. Asymptotic normality of statistics used for classification is proved. Results of numerical modeling are presented.



9.
Comparison of feldkamp algorithm with fourier synthesis algorithm for 3d tomography

A. V. Likhachev
Novosibirsk
Pages: 77-88

Abstract >>
A new slice-by-slice reconstruction algorithm based on the Fourier synthesis method is proposed for projection geometry with a source moving along a circle. The time estimate of reconstructing by this algorithm on an N x N x N grid from M projections is O(MN2log2N). The estimate is O(N / log2N) times less than that for classical implementation of Feldkamp algorithm. A computational experiment showed that the Feldkamp algorithm yields in accuracy to the method of Fourier slice-by-slice synthesis in the case of a small radius of the source movement circle (3



10.
The tracking laser doppler system based on optimal instantaneous frequency estimates

V. S. Sobolev, A. A. Stolpovskii, A. M. Shcherbachenko, G. A. Kashcheeva, and E. N. Utkin
Novosibirsk
Pages: 89-100

Abstract >>
New approaches to designing highly accurate laser Doppler measurement systems (LDMS) are presented. They are based on the concept of instantaneous frequency of the analytical Doppler signal, the algorithms of its maximum likelihood estimates, and the principles of tracking filtering. LDMS based on those principles is described and results of its metrological investigation and applications in an aerodynamic experiment are presented. The adopted engineering solutions are presented and substantiated.



11.
Evolutionary synthesis of algorithms based on templates

O. G. Monakhov
Novosibirsk
Pages: 101-109

Abstract >>
A new algorithm synthesis approach based on evolutionary computation, templates and skeletons of algorithms, and a given set of pairs of input-output data is described. An evolutionary synthesis algorithm combining advantages of genetic algorithms and genetic programming, and used for automation of discovery and synthesis of some computational and combinatorial algorithms is presented.