Madiega, Blaise, Olivier, Mathieu, Alahassa, Konrad Tagnon Amen et Brou, Christ-Marie Anzian.
2025.
« Enhancing noisy PIV measurements through signal processing and machine learning techniques ».
In Proceedings of the CSME-CFDSC-CSR 2025 International Congress (Montreal, QC, Canada, May 25-28, 2025)
Coll. « Progress in Canadian Mechanical Engineering », vol. 8.
Prévisualisation |
PDF
234 - Enhancing noisy PIV measurements t.pdf - Version publiée Licence d'utilisation : Tous les droits réservés aux détenteurs du droit d'auteur. Télécharger (1MB) | Prévisualisation |
Résumé
Particle Image Velocimetry (PIV) is a cornerstone technique in experimental fluid dynamics, yet its accuracy and reliability can be severely affected by noise and insufficient resolution in the measured velocity fields. In this work, we propose a deep learning framework that integrates signal processing methods with a hybrid U-Net-LSTM convolutional neural network to enhance data quality and resolution. Our approach applies filtering techniques for noise reduction and leverages learned feature extraction for improved particle segmentation, ultimately boosting measurement fidelity. We evaluate this pipeline using a specialized PIV-UQ dataset containing raw images (PIV-MS, low resolution measurement system) and high-fidelity reference measurements from a stereoscopic system (PIV-HDR). Results indicate that effective denoising in conjunction with the U-Net-LSTM architecture significantly refines pixel-level velocity estimations. (Code available at https://github.com/Bmadios/MLSP4PIV_enhancing)
| Type de document: | Compte rendu de conférence |
|---|---|
| Éditeurs: | Éditeurs ORCID Hof, Lucas A. NON SPÉCIFIÉ Di Labbio, Giuseppe NON SPÉCIFIÉ Tahan, Antoine NON SPÉCIFIÉ Sanjosé, Marlène NON SPÉCIFIÉ Lalonde, Sébastien NON SPÉCIFIÉ Demarquette, Nicole R. NON SPÉCIFIÉ |
| Date de dépôt: | 18 déc. 2025 15:16 |
| Dernière modification: | 18 déc. 2025 15:16 |
| URI: | https://espace2.etsmtl.ca/id/eprint/32450 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |