High-precision wavefront reconstruction from a single-shot interferogram using a physics-driven hybrid feature...

Area of Science:

• Optics and Photonics
• Metrology
• Artificial Intelligence in Science

Background:

• Interferometric wavefront reconstruction is crucial for precision optical metrology.
• Traditional methods face challenges in phase extraction and error accumulation.
• Existing deep learning approaches may lack physical constraints and introduce unwrapping errors.

Purpose of the Study:

• To develop a novel method for high-precision wavefront extraction from single-shot interferograms.
• To address limitations of traditional and current deep learning-based wavefront reconstruction techniques.
• To introduce a physics-driven hybrid feature calibration network (HFC-Net).

Main Methods:

• A physics-driven hybrid feature calibration network (HFC-Net) was designed and optimized.
• The HFC-Net utilizes a single-shot interferogram for wavefront extraction.
• Performance was evaluated using simulations and experimental data.

Main Results:

• HFC-Net achieved very low root mean square (RMS) errors: 0.00031λ on simulated data and 0.0021λ on real data.
• The processing time was efficient at 87.14 ms.
• The method demonstrated superior efficiency compared to Fourier transform and other deep learning models.

Conclusions:

• The proposed HFC-Net enables accurate and efficient single-shot interferometric wavefront reconstruction.
• This physics-driven approach overcomes key limitations of existing methods.
• The HFC-Net is valuable for applications like phase object and dynamic aberration measurement.