Abstract: In the field of astronomy, the classification of galaxies has always been a hot and difficult problem. In recent years, some scholars have applied machine learning to the simple classification task of galaxy morphology, but in the process of classification, there are a series of problems, such as feature selection difficulty, feature omission, classifier selection difficulty and so on. Galaxies can be roughly divided into elliptical galaxies, spiral galaxies, lenticular galaxies and irregular galaxies in visual morphology. In this paper, GMC (Galaxy morphological classification) which is a more accurate classification method is proposed for the photometric images of galaxies in SDSS DR16, Galaxy Zoo2 and EFIGI catalog. Firstly, we cut and denoise the images, and use rotation, translation, scaling and other methods to enhance the data. Finally, we build the GMC-net to classify photometric images. According to the classification results, the classification accuracy of spiral galaxies, elliptical galaxies, lenticular galaxies and irregular galaxies in different databases are 98.29%, 98.49%, 99.18% and 99.91%, respectively; The average classification accuracy of four different galaxies from the same database EFIGI catalog is 99.34%. The experimental results show that GMC performs better than other classification methods, and can be used to classify galaxies more effectively.