Efficient classification of orbital angular momentum superposition modes based on multimodal feature fusion

Deep learning-based classification of orbital angular momentum (OAM) modes is important for OAM-multiplexed communication systems, yet the multimodal classification of superimposed OAM modes remains rarely explored. We propose a classification method that fuses polar-coordinate features with two-dimensional features extracted from local and global branches for atmospheric transmission of superimposed OAM modes. A polar coordinate transform converts the two-dimensional light intensity distribution into one-dimensional radial and angular sequences. Meanwhile, the image branch captures local structural details and global semantic information. The proposed multimodal feature fusion method provides complementary radial-angular and image-domain information for classification. Experiments show that the proposed model achieves competitive classification accuracy with a clear convergence advantage over the compared baseline classifiers, demonstrating the effectiveness of multimodal fusion for superimposed OAM mode recognition under atmospheric turbulence.