Orbital angular momentum mode recognition under ocean turbulence channel by DCNN-RF model based on Adma optimization

Abstract

Gaussian beam carries orbital angular momentum with infinite Hebert space distribution, which is expected to realize large-capacity and long-distance transmission of underwater wireless optical communication. To address the problem that Gaussian beam will be dispersed by diffusion and crosstalk after passing through an ocean turbulence channel, resulting in blurring of OAM mode features and difficulty in extracting OAM mode information, this paper proposes a DCNN-RF mode recognition model, which aims to extract the multidimensional feature information of the RGB and HSV of the OAM mode image through a dual-channel CNN, and then optimize the CNN model by combining with Adam optimizer parameters, and finally the extracted multidimensional features are fused by RF classifier and classify the OAM mode. The experimental results show that the DCNN-RF model achieves 100% recognition rate for weak and medium turbulence, and 97% to 99.5% recognition rate for strong turbulence environment. Analyzing the recognition rate analysis of CNN, CNN-SVM, ResNet50-SVM model and DCNN-RF model, DCNN-RF model has more excellent OAM mode detection ability, and the recognition rate of DCNN-RF model reaches 99.5% in the actual OAM experimental system. This will provide a new reference for mode recognition in OAM optical communication.

Keywords

• Orbital angular momentum
• Wireless optical communication
• Mode recognition
• Dual-channel convolutional neural network
• Random forests