Robust Wavefront Sensing: Experimental Demonstration of Intensity and Slopes Neural Network (ISNet)

Abstract

This final report discusses experimental results of the intensity and slopes neural network (ISNet) for turbulent wavefront reconstruction using Shack-Hartmann data. The network was previously trained on simulated data, and transfer learning is applied to operate on benchtop experimental setup. Ground truth data was obtained using a deformable mirror to generate known turbulent wavefronts. After transfer learning, ISNet is tested on unknown wavefronts using a phase screen in the signal path. Results show improved wavefront reconstruction from ISNet compared to Shack-Hartmann reconstruction. A parallel digital holography wavefront sensor is also used to verify the results.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 22, 2022
Accession Number
AD1189490

Entities

People

  • Harshil Dave

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Adaptive Optics
  • Algorithms
  • Arrays
  • Artificial Intelligence Computing
  • Artificial Intelligence Software
  • Convolutional Neural Networks
  • Data Sets
  • Deep Learning
  • Deformable Mirrors
  • Demonstrations
  • Department Of Defense
  • Detectors
  • Distortion
  • Experimental Data
  • Information Operations
  • Information Science
  • Intensity
  • Learning
  • Machine Learning
  • Machines
  • Measurement
  • Neural Networks
  • Optics
  • Physics
  • Turbulence
  • Wavefronts

Fields of Study

  • Physics

Readers

  • Atmospheric Science / Meteorology, specifically Wind Wave Turbulence.
  • Image Processing and Computer Vision.
  • Neural Network Machine Learning.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Neural Networks