In recent years, the number of space objects located in near-Earth outer space, especially in the near operational zone, has increased significantly due to the build-up of space groupings, including dual-use (for example, Starlink) and the remnants of their vital activity (space debris). This factor increases the importance of the task of recognizing and classifying space objects by type in the shortest possible time and entering them into the main catalog of space objects. The developed methodology allows automated analysis of optical images of space objects using software to solve the problem of their recognition and classification by type using a convolutional neural network. The purpose of the study is to increase the efficiency of processing and analysis of optical images of spacecraft. The experimental results of the study confirm the achievement of the research goal. The developed methodology contributes to the development of software and hardware for image processing and can be used in calculations and data preparation for information support of interested officials. For the first time, a training set for a convolutional neural network has been prepared using real optical images of spacecraft obtained in the visible range.