Multichannel left-subtract-right feature vector piston error detection method based on a convolutional neural...

Area of Science:

• Astronomy and Astrophysics
• Optical Engineering
• Machine Learning

Background:

• Large-scale synthetic-aperture telescopes require high-precision co-phasing for optimal performance.
• Traditional piston error detection methods face challenges in range and noise resistance.

Purpose of the Study:

• To propose a novel multichannel piston error detection method for synthetic-aperture telescopes.
• To enhance the accuracy and detection range of co-phasing adjustments.
• To address the limitations of traditional neural network training datasets.

Main Methods:

• A convolutional neural network (CNN) based multichannel left-subtract-right (LSR) feature vector piston error detection method.
• Development of a scheme for constructing large training datasets for the CNN.
• Simulation-based verification of the proposed method's performance.

Main Results:

• The proposed method achieves a wide detection range of (-139λ, 139λ) with λ = 720 nm.
• High precision and strong noise resistance inherited from the DFA-LSR method.
• Guaranteed accuracy of at least 94.96% with large samples.
• Root mean square error of 10.2 nm at a signal-to-noise ratio of 15.

Conclusions:

• The CNN-based multichannel LSR method offers a robust solution for large-scale, high-precision co-phasing adjustment of synthetic-aperture telescopes.
• The proposed dataset construction scheme overcomes traditional limitations in training data acquisition.
• This method significantly advances the capabilities for precise optical alignment in astronomical instrumentation.