Related Experiment Video
Updated: Aug 16, 2025
16:11
Implementation of a Reference Interferometer for Nanodetection
Published on: April 26, 2014
9.4K
Fringe Detection and Displacement Sensing for Variable Optical Feedback-Based Self-Mixing Interferometry by Using
Asra Abid Siddiqui1, Usman Zabit1, Olivier D Bernal2
1School of Electrical Engineering and Computer Science, National University of Sciences and Technology (NUST), Islamabad 44000, Pakistan.
Sensors (Basel, Switzerland)
|December 23, 2022
Summary
Deep neural networks enhance laser feedback self-mixing interferometry (SMI) for displacement sensing. This overcomes challenges from variable optical feedback and speckle, improving accuracy for remote targets.
Area of Science:
- Optics and Photonics
- Machine Learning Applications
- Metrology and Measurement Science
Background:
- Self-mixing interferometry (SMI) using laser feedback offers potential for precise displacement sensing.
- Commercialization is hindered by performance degradation under variable optical feedback, particularly with non-cooperative targets and speckle.
- Existing methods struggle with reliable fringe detection in noisy, speckle-affected SMI signals.
Purpose of the Study:
- To develop and evaluate deep neural networks for robust interferometric fringe detection and displacement measurement in SMI systems.
- To address the challenge of variable optical feedback and speckle noise in laser feedback-based sensors.
- To propose an automated method for labeling SMI fringes to create large training datasets for machine learning models.
Main Methods:
- Training deep neural network models (Yolov5 and EfficientDet) on experimental SMI signals with variable optical feedback.
- Implementing an automatic fringe labeling technique to generate extensive training data.
- Quantifying network performance using metrics such as fringe detection accuracy, signal-to-noise ratio, depth of modulation, and execution time.
Main Results:
- Demonstrated successful fringe detection and displacement measurement using deep learning models under challenging optical feedback conditions.
- Achieved reliable performance across different laser-diode sensors and varying noise/speckle environments.
- Validated the effectiveness of the automatic fringe labeling method for dataset generation.
Conclusions:
- Deep neural networks, specifically Yolov5 and EfficientDet, significantly improve the robustness and accuracy of self-mixing interferometry displacement sensing.
- The proposed automated labeling method facilitates the creation of large datasets, crucial for training effective machine learning models in this domain.
- This work paves the way for wider commercial adoption of SMI sensors in applications involving non-cooperative targets and variable optical feedback.
