Centroid computation for Shack-Hartmann wavefront sensor in extreme situations based on artificial neural networks

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Astronomy

Background:

• Shack-Hartmann wavefront sensors (SHWFS) are crucial for adaptive optics (AO) systems.
• Traditional centroid calculation methods struggle with environmental light and noise pollution.
• Robust wavefront sensing is essential for high-resolution imaging and astronomical applications.

Purpose of the Study:

• To develop a novel method for accurate centroid calculation in SHWFS under adverse environmental conditions.
• To enhance the robustness and reliability of SHWFS in low signal-to-noise ratio (SNR) scenarios.
• To demonstrate the effectiveness of artificial neural networks for wavefront sensing challenges.

Main Methods:

• A specialized artificial neural network, SHWFS-Neural Network (SHNN), was designed for SHWFS.
• The method transforms spot detection into a classification problem to identify the spot center.
• SHNN models with varying hidden layer neurons (SHNN-50, SHNN-900) were evaluated.

Main Results:

• SHNN-900 achieved a 0% False Rate in extreme low SNR (peak SNR=3), compared to 26% for traditional methods.
• SHNN-900 maintained a 0% False Rate with increased environmental light interference.
• Wavefront reconstruction showed a significant decrease in Root Mean Square (RMS) residual from 0.5349 um to 0.0383 um.

Conclusions:

• The proposed SHNN method significantly improves SHWFS performance and robustness in polluted environments.
• SHNN offers a superior alternative to traditional centroid calculation methods under challenging conditions.
• This advancement has implications for improving adaptive optics system capabilities.