Aerosol characterization on noisy surfaces by deep-learning methods
Résumé
Experimental studies to characterize aerosol deposition
in flow system (such as ventilation duct) may require
avoiding a sampling substrate which modifies the flow in
the boundary layer, influencing the deposition or the
interaction forces. However, measuring the deposition of
aerosols on surface without any intermediate substrate
presents detection difficulties (Costa et al. 2021) which
can be overcome by direct visualization of aerosols by
microscopy, using the microscope as a scanner.
However, analysing the deposition of aerosols on any
surfaces by microscopy generates certain constraints
regarding the use of known visualization techniques:
inhomogeneous image backgrounds due to the surface
structure, unsharp images due to the non-flatness of the
real considered surfaces. Common image processing is
then not very effective.
The objective of this work is therefore to deploy
advanced computer vision methods to process images
acquired on typical surfaces of ventilation ducts
(stainless steel).
Little work exists to date on the detection of aerosols
using neural networks. Monchot et al. (2021) were
interested in nanoparticles, as were Merouane et al.
(2023) but in these works, the particles stand out
relatively well from the background.
Origine | Fichiers produits par l'(les) auteur(s) |
---|---|
licence |
Copyright (Tous droits réservés)
|