Abstract. The paper considers the problem of recognition of a given class of objects by the result of multispectral measurements (spectral brightness of signals) and available spectral and statistical characteristics of the given classes. On the basis of probabilistic and statistical considerations, as well as quantization, the heuristic recognition criterion is proposed. Based on the criterion, the heuristic method of recognition is presented. Modifications of the method are proposed to improve its reliability and efficiency.

Keywords: class of recognition objects, density distribution, normal distribution, multivariate distribution density, dispersion, standard deviation, discrete distribution, alternative hypothesis, heuristic recognition criteria.