Optical image centroid prediction based on machine learning for laser satellite communication.

Area of Science:

• Optical engineering
• Machine learning
• Satellite communication

Background:

• Optical image tracing is crucial for satellite-to-ground laser communication.
• Atmospheric turbulence significantly impacts laser communication link stability.
• Machine learning offers powerful tools for modeling complex, nonlinear systems.

Purpose of the Study:

• To develop a novel machine learning model for predicting centroid positions in optical image tracing.
• To enhance the real-time performance and accuracy of satellite-to-ground laser communication systems.
• To account for physical characteristics of light fields on CMOS sensors.

Main Methods:

• A hybrid model integrating a Convolutional Neural Network (CNN) module and a Long-Short Term Neural Network (LSTM) module was developed.
• The model incorporates a preprocessing module to handle digital images in time series.
• Physical situations relevant to light fields on CMOS sensors were integrated into the model architecture.

Main Results:

• The developed model accurately predicts centroid positions in the presence of atmospheric turbulence.
• Real-time prediction capabilities were achieved, suitable for practical laser satellite communication applications.
• The model demonstrates improved performance compared to traditional neural network approaches by incorporating physical considerations.

Conclusions:

• The hybrid CNN-LSTM model provides a robust solution for real-time optical image tracing in satellite laser communication.
• Accurate centroid position prediction under atmospheric turbulence is achievable with advanced machine learning techniques.
• This approach facilitates more reliable and efficient satellite-to-ground laser communication.