Multidimensional Information Assisted Deep Learning Realizing Flexible Recognition of Vortex Beam Modes

Due to countless orthogonal eigenstates, light beams with orbital angular momentum(OAM) have a large potential information capacity. Recently, deep learning has been extensively applied in recognition of OAM mode. However, previous methods require constant distance between laser and receiver. The accuracy will drop quickly if the testing set deviates from training set. Previous also difficulty distinguishing modes positive negative topological charges. In order further exploit huge dimension state space, we proposed multidimensional assisted flexible (MIADLFR) method make use both intensity spectrum for first time achieve unlimited by sign TC high accuracy. Also, MIADLFR can reduce computational complexity significantly requires much smaller