10.2514/6.2026-2248">
 

Comparison of Methods for Wavelength Determination from Laser Speckle Patterns

Document Type

Conference Proceeding

Publication Date

1-2026

Abstract

Sensing the wavelength of laser radiation is a challenging task which has scientific and military applications. These applications include laser calibration, medical imaging as well as industrial quality control in the manufacture of laser cavities. Currently, different kinds of sensors can be used to estimate the wavelength of light present in a laser beam but generally fall into either the category of diffraction gratings or Michaelson interferometers. These kinds of sensors are severely limited in the range of wavelengths they can measure. For example, diffraction gratings have a blaze angle that is chosen to maximize the efficiency of the grating at concentrating light at a chosen wavelength into the primary diffraction order. Other wavelengths appear in other orders with a lower overall efficiency. Michaelson interferometers do not suffer the same limitations as diffraction gratings but are generally expensive to build and maintain and possess low modulation efficiencies at shorter wavelengths where wavefront errors between the arms of the interferometer become large as wavelength decreases. An emerging method for sensing wavelength involves the use of a diffusive optic, such as shower glass or multimode fibers which produce speckle patterns that can be captured by detector arrays and analyzed to determine the wavelength of the light that generated them. Shower glass is cheap, durable and produces speckle patterns over the entire ultraviolet to long wave infrared spectrum. There exist two current methodologies used to calculate the wavelength from speckle pattern including speckle correlation or a machine learning model. In the correlation method, the laser speckle patterns are correlated with pre-measured patterns representing different wavelengths. By comparing the correlation value between an unknown speckle pattern and known speckle patterns, it is expected that known speckles of a similar or exact wavelength match will have a statistically significant correlation when compared to wavelengths of a farther range away from the unknown. In the machine learning method, a model is formed and trained on labeled speckle patterns. After training, it is expected that the model will detect the small variance and pattern differences that occur as wavelength changes when a speckle pattern is created. It is expected that the model will be able to report with a high degree of accuracy the wavelength used to generate a speckle pattern when provided a speckle pattern of unknown wavelength within its training range. Both of the aforementioned techniques for laser wavelength determination require training data which potentially limits the utility of the sensor as environmental changes or hardware degradation could cause the system to change in some way that is inconsistent with its behavior during the training cycle. This paper considers a new approach that does not require training but instead relies on a statistical model for laser speckle. The new method is tested for accuracy and precision against the existing methods on simulated laser speckle data sets. The results show that the new method achieves better accuracy and precision than the machine learning approaches over a large portion of the visible spectrum for the simulated data used in this study.

Comments

The full conference paper is accessible by subscription or purchase from AIAA using the DOI link below.

Conference Session: Novel Sensors, Algorithms, and Sensing Applications

Source Publication

AIAA SCITECH 2026 Forum

This document is currently not available here.

Share

COinS