Session: Ultrasonic Arrays I

Paper Number: 170877

170877 - In-Situ Microplastic Detection/sizing Using Circular Ultrasonic Arrays With Ultrasonic Imaging and Deep Learning Image Segmentation. 

Abstract: 

The widespread presence of microplastics in our oceans presents significant threats to marine ecosystems and potentially for human health, demanding innovative detection and monitoring solutions. The National Research Council Canada's microplastic project is pioneering the development of a state-of-the-art in situ instrument for real-time detection, sizing, and identification of microplastic particles. This presentation will showcase the project's overall framework, including various sensing technologies such as spectroscopy, and will emphasize advancements in ultrasonic imaging and deep learning methodologies.

A key component of this initiative is the deployment of circular ultrasonic arrays, which significantly enhance the ability to capture detailed acoustic images of microplastics. Techniques like Total Focusing Method (TFM) provide high-quality reconstructions but are data-intensive and slow, whereas single-Circular Wave Imaging (CWI) enables faster data acquisition, albeit with increased image noise and artifacts. We will demonstrate image reconstructions using these methods from both simulated and experimental data within this circular geometry, addressing challenges related to the geometry and probe wedge coupling.

Our research also focuses on leveraging deep learning models, specifically a U-net architecture to segment particles in CWI and TFM images while potentially reducing the number of required sources for TFM. By training this model on synthetic datasets generated with the j-Wave Python library made for acoustic simulation, we aim to achieve high-quality imaging with reduced power consumption and enhanced real-time processing capabilities, facilitated by rapid acquisition techniques. Deploying these models on edge computing platforms further enhances the practicality of this technology for on-site monitoring.

Our project highlights the transformative potential of integrating cutting-edge sensor technology with AI-driven analytics to tackle the global challenge of microplastic pollution, paving the way for more effective monitoring and remediation strategies.

Presenting Author: Christophe Bescond National Research Council Canada

Presenting Author Biography: Christophe Bescond has obtained his PhD in 1997 from the University of Bordeaux (France) in collaboration with Tongji University in Shanghai (China). He then joined the Laboratory of Photo-Acoustics and Laser-Ultrasound in the Engineering Physics department at the École Polytechnique de Montréal in Quebec (Canada) for a postdoctoral fellowship and then a research associate position until 2000. Christophe Bescond was then a research associate at the NRC Industrial Materials Institute (IMI) from 2000 to 2005. After 2005, he pursued a career in various industrial sectors such as nuclear, aerospace, petrochemical, Steel industry, manufacturing and others as NDT specialist. From 2009 to 2013, at Bombardier Aerospace, he was a specialist in advanced ultrasound for the development of new programs and for NDT R&D projects. Since 2013, he is back at the NRC where his expertise is the development of ultrasonic and photoacoustic sensors for mining, environmental, aerospace and other applications. From 2015 to 2023, he was leader of the ultrasonic technologies team. Currently, he supervises a major project and development for the in-situ detection of microplastics in the oceans.

Authors:

Christophe Bescond National Research Council Canada
Jean-Hughes Fournier-Lupien National Research Council Canada