High-robustness underwater vortex beam recognition using conjugate superimposed OAM modes and a deep residual network

Area of Science:

• Optical communication
• Underwater sensing
• Machine learning for signal processing

Background:

• Orbital Angular Momentum (OAM) modes in underwater optical communication are susceptible to distortion from scattering, turbulence, and occlusions.
• Existing methods struggle to maintain reliable OAM mode recognition in challenging underwater environments.

Purpose of the Study:

• To propose and experimentally validate a robust OAM mode recognition scheme for underwater optical communication.
• To enhance the resilience of OAM communication systems against environmental perturbations.

Main Methods:

• Utilized conjugate superimposed OAM beams to preserve structural information.
• Employed a deep residual network (ResNet-50) for mode classification.
• Simulated nine underwater disturbance scenarios by varying kaolin concentration, water pump power, and introducing occlusions.

Main Results:

• Achieved approximately 100% classification accuracy for 16 OAM modes across all tested disturbance levels.
• Demonstrated reliable recognition even with occlusions covering up to 50% of the beam cross-section.
• The proposed scheme showed high generalization capability without overfitting.

Conclusions:

• The combination of conjugate superimposed OAM beams and ResNet-50 provides a practical and intelligent demodulation framework.
• This approach significantly improves the robustness of underwater OAM communication systems.
• The findings pave the way for deploying reliable OAM communication in real-sea conditions.