Atmospheric Turbulence Study with Deep Machine Learning of Intensity Scintillation Patterns

A new paradigm for machine learning-inspired atmospheric turbulence sensing is developed and applied to predict the atmospheric turbulence refractive index structure parameter using deep neural network (DNN)-based processing of short-exposure laser beam intensity scintillation patterns obtained with...

Full description

Saved in:
Bibliographic Details
Published in:Applied sciences 2020-11, Vol.10 (22), p.8136
Main Authors: Vorontsov, Artem M., Vorontsov, Mikhail A., Filimonov, Grigorii A., Polnau, Ernst
Format: Article
Language:English
Subjects:
Accuracy
active imaging
adaptive optics
Analysis
atmospheric remote sensing
Atmospheric turbulence
Computer simulation
Datasets
directed energy
Evaluation
Experiments
free-space laser communication
Hypotheses
Information processing
Laser beams
Lasers
Learning algorithms
lidars
Machine learning
Methods
Neural networks
Optics
Propagation
Refractivity
Scintillation
Sensors
Signal processing
Temporal resolution
Testing
Turbulence models
Wave propagation
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A new paradigm for machine learning-inspired atmospheric turbulence sensing is developed and applied to predict the atmospheric turbulence refractive index structure parameter using deep neural network (DNN)-based processing of short-exposure laser beam intensity scintillation patterns obtained with both: experimental measurement trials conducted over a 7 km propagation path, and imitation of these trials using wave-optics numerical simulations. The developed DNN model was optimized and evaluated in a set of machine learning experiments. The results obtained demonstrate both good accuracy and high temporal resolution in sensing. The machine learning approach was also employed to challenge the validity of several eminent atmospheric turbulence theoretical models and to evaluate them against the experimentally measured data.
ISSN:2076-3417
2076-3417
DOI:10.3390/app10228136