Problem solving in astronomy, using computer methods is a very topical issue nowadays. The topicality of object classification problem has been increasing during the last couple of years, especially because of the radio signal received from cosmos, the growth of the unclassified images from the telescope Hubble and the activity of such projects as Sloan Digital Sky Survey. Such projects as Sloan Digital Sky Survey provide enormous amounts of digital images from all the sides of the Universe. Today, the problem of galaxies classification is being solved using manual data classification, but it can be solved much more efficiently with the help of computer methods. Nowadays, there is a range of theories in astronomy, which could be proved or disapproved, provided that particular information, concerning the evolution of galaxies, is available. In order to obtain this information, astronomers investigate large groups of objects, belonging to the same class and existing at the different stages of development. Galaxies get born and die, but not in one day. These processes are very slow, and the current number of the starry sky images is enormous. Today, even quite large groups of volunteers cannot classify all the images of galaxies on the existing photographs. Therefore, astronomy needs the help of information technologies and computer methods, which are already applied successfully in other scientific fields, such as biology and engineering. Taking into account the capacity of computer’s memory and performance of modern computers, the problem of the analysis of huge amount of imagery data can be solved. THE OBJECT OF THE RESEARCH: computer methods of statistical astronomy. THE SUBJECT OF THE RESEARCH: Object recognition algorithms, aimed at galaxy classification. THE HYPOTHESIS OF THE RESEARCH: Computer system parameters can be adapted to enable automatic galaxy classification with the average accuracy, corresponding to the same or higher level of accuracy in comparison with the manual data processing. THE AIM OF THE RESEARCH: To broaden the application of computer systems, aimed at image recognition in bio-informatics, adapting them to the scientific tasks of astronomy. RESEARCH RESULTS: The ET-BOF method was selected for the experiment and adapted for the tasks of the research (method has been designed at The University of Liège in Belgium). The software configuration, training and testing was performed from November 2009 to April 2010. The best results of automatic recognition were achieved when using image segmentation with colour threshold and convolution matrix method and applying these methods for the etalon set of galactic images. The results were compared with the data from the international Galaxy Zoo open project. The experiment has proved that automatic classification of galaxies ensures the same or higher accuracy results in comparison with the manual classification. The experiment with the great set of data has proved that the automatic galaxy classification was performed with the accuracy of at least 90%. The existence of the 10% error in the automatic classification is not significant in the frame of the research. Fundamental astronomy is interested in information concerning types of galaxies in large clusters. The achieved level of accuracy is considered to be acceptable for creation of the Universe evolution models on a large scale. Thus, it is possible to say that the aim of the research has been achieved and the hypothesis has been proved.