I. A. Sinitsyn, V. V. Tatarinov, V. N. Gavryshin

Parametric identification and definition and characteristics of reflecting object to some class is relevant in specle-interferonometric non-destructive control problem. Target setting is done in standard terms of statistic solution theory and expression for likelihood function is obtained, which is sufficient statistics for Bayesian and non- Bayesian decision rules. It is shown that calculations of likelihood function meanings come across con-siderable problems. Spectral analysis of real speckle-optical object images has shown that spectral concentration power may be approximated by fractal-rational facilities of spatial frequencies. This makes it possible to obtain extremely constructive mathematical image model as an analog of Markov sequence. Parameters of approximating of fractal-rational spectral concentrations, obtained by the smallest squares method may be used for solving classification problem. The results of experimental data processing confirm workability of the offered algorithm.