Modeling and recognition of complex interference patterns in multi-surface phase-shifting interferometry.

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Metrology

Background:

• Traditional multi-surface phase-shifting interferometry (MS-PSI) algorithms face challenges with phase-shift accuracy and experimental variability.
• Existing methods are sensitive to environmental factors and calibration errors, limiting measurement precision.

Purpose of the Study:

• To develop a robust and accurate deep learning framework for multi-surface phase-shifting interferometry.
• To overcome the limitations of traditional MS-PSI algorithms regarding phase-shift sensitivity and experimental conditions.

Main Methods:

• Proposed MSAI-Net (multi-scale attention interference network) combining multi-scale convolutional feature extraction.
• Integrated SE attention and hierarchical LSTM-self-attention modules for enhanced fringe recognition.
• Developed a simulated dataset with nonlinear wavelength modulation and higher-order interference, discretizing phase labels into 500 classes.

Main Results:

• MSAI-Net achieved 96.15% accuracy in simulations.
• Demonstrated robustness against phase-shift errors and higher-order interference.
• Experimental validation showed low RMS error even with ±25% phase deviations.

Conclusions:

• MSAI-Net offers a reliable and highly accurate solution for multi-surface interferometric measurements.
• The proposed network effectively addresses the sensitivity issues of traditional MS-PSI algorithms.
• This deep learning approach advances the field of precision optical metrology.