|
Guest >> Sign In |
|
|
DIRECTED
ENERGY
PROFESSIONAL
SOCIETY
Abstract: 25-Symp-007
|
|
|
|
UNCLASSIFIED, PUBLIC RELEASE
Machine learning for single-image amplitude, wavefront, and polarization sensing
Amplitude, wavefront, polarization, and wavelength (spectral amplitude) are four basic parameters of a laser beam. The correct characterization of the spatial profile of these parameters plays a critical role not only for the development of high-power lasers but also for predicting their propagation to the final target. Commercial products are available to retrieve each spatial profile but with separate devices. For example, conventional camera sensors exist for amplitude profile detection, and a Shack-Hartmann wavefront sensor can perform wavefront profile detection. This paper studies the potential of using machine learning to retrieve the spatial distribution of amplitude, wavefront and polarization from one camera image. Spatial information for three laser parameters (amplitude, wavefront, and polarization) is encoded into one camera image by placing an optical encoder a few millimeters in front of a camera sensor. Each spatial profile can be retrieved from the encoded camera image. As one example, we studied the performance of amplitude, wavefront, and polarization profile retrieval from using two Convolutional Neural Network (CNN) architectures compared to that from using conventional algorithms.
©2025 The MITRE Corporation. ALL RIGHTS RESERVED.
Approved for Public Release; Distribution Unlimited. Public Release Case Number 25-0539
UNCLASSIFIED, PUBLIC RELEASE
|