Oceanic turbulence parameters recognition based on convolutional neural network

The distortion induced by ocean turbulence has a substantial impact on the propagation of light in water, posing challenges for applications including underwater wireless optical communications and submarine surveys. Obtaining accurate information about the properties of oceanic turbulence (OT), particularly the parameters describing OT, is crucial for addressing these challenges and enhancing the performance of such applications. In this paper, we propose a convolutional neural network (CNN) and validate its ability to recognize OT parameters. The physical quantities of oceanic turbulence collectively influence the formation and strength of turbulence. We recognize the dissipation rate of temperature variance chi T and the turbulent kinetic energy dissipation rate epsilon, taking into account various balance parameter omega, transmission distance z. Furthermore, in order to simultaneously recognize chi T and epsilon, we enhanced the existing network by modifying the output structure, resulting in a dual-output architecture that facilitates concurrent classification of both chi T and epsilon. Our method for classifying turbulence parameters will contribute to the field of underwater wireless optical communication and promote its further development.