Resumen
The authors continue to study the application of machine learning methods to Geophysics problems. The focus of this work was the procedure for selecting frequencies for RGB-mixing. Previous work of the authors in this direction used a heuristic approach to the selection of frequencies for RGB-mixing. A key issue in RGB rendering is the choice of frequencies to mix. To date, there is no reasonable approach to determine the optimal combination of frequency characteristics to obtain the most informative result. At the same time, the evaluation of this information content is subjective, which also makes it difficult to choose a method for solving this problem.The article considers several approaches to the selection of frequency characteristics to obtain optimal RGB-representation of the results of spectral decomposition. Achieving the optimal choice is based on various prerequisites: subjective assessment (this approach is called "heuristic"), the distribution of reservoir thicknesses (this approach is called "geological") and the total amplitude-frequency characteristics of the wavelets used (this approach is called "optimization").In conclusion, the results of the practical application of the described techniques are presented, as well as the advantages of the choice of frequency characteristics by solving the optimization problem. At the same time, the authors note the absence of quantitative criteria for the evaluation of different approaches, since the use of any of the techniques allows to solve the main problem of RGB-analysis (i.e., to identify geological objects in the wave field).Thus, the authors examined two well-known approaches to the choice of frequencies for carrying out an RGB mixing and proposed a new method based on the solution of the optimization problem, allowing you to get away from the subjective choices of the frequencies involved in the formation of a false color representation of the results of spectral decomposition.