OAM Pattern Recognition Method Based on ResNet-18 Migration Learning

Aiming at the application of Orbital Angular Momentum (OAM) in the field of optical communication, this paper proposes an OAM pattern recognition method based on ResNet-18 transfer learning. The OAM mode is a light wave with a spiral phase structure, which has the characteristics of high degree of freedom and large capacity, and has broad application prospects in the field of optical communication. However, detection and classification of OAM modes is difficult due to their complex wavefront structures. To address this issue, we use a pretrained ResNet-18 model to classify OAM patterns. This method uses the pre-trained ResNet-18 model for feature extraction and fine-tuning, which can effectively improve the classification performance of the model. Through experimental verification, the method has high accuracy and stability in the classification task of OAM mode. The research in this paper provides strong support for the application of OAM technology, and has certain reference value for image classification problems in other fields.