A Precise and Fast Droplet Parameter Inversion Algorithm for Rainbow Scattering Detection

The rainbow scattering technique holds significant promise for simultaneously detecting micrometer-sized droplet size and refractive index. Reported here is a fast and lightweight deep-learning-based rainbow inversion algorithm containing a unique signal preprocessor and a novel signal inversion net...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE sensors journal Jg. 24; H. 19; S. 30608 - 30618
Hauptverfasser: Li, Tianchi, Li, Can, Li, Ning
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.10.2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Accuracy
Algorithms
Deep learning
droplet size
Droplets
Fitting
inversion
Machine learning
Multilayer perceptrons
Optical scattering
Optical variables control
Parameters
rainbow scattering
Rainbows
Refractive index
Refractivity
Scattering angle
ISSN:1530-437X, 1558-1748
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The rainbow scattering technique holds significant promise for simultaneously detecting micrometer-sized droplet size and refractive index. Reported here is a fast and lightweight deep-learning-based rainbow inversion algorithm containing a unique signal preprocessor and a novel signal inversion network. Rainbow signal is preprocessed based on scattering intensity, scattering angle, and its intensity-angle area to effectively preserve the physical features. A multilayer perceptron (MLP)-based rainbow inversion network is constructed for more effective inversion of droplet size and refractive index at the global sensory field level. The algorithm achieves a 50fold speedup compared with conventional methods without compromising accuracy. Extensive simulations demonstrate an average relative error of <inline-formula> <tex-math notation="LaTeX">{0}.{89}\% </tex-math></inline-formula> in droplet size estimation and an average absolute error of <inline-formula> <tex-math notation="LaTeX">{3}.{81}\times {10} ^{-{5}} </tex-math></inline-formula> in refractive index determination. Experimental validation further confirms the algorithm's reliability. This work not only offers significant improvements in computational speed and accuracy but also opens new avenues for advanced droplet parameter inversion algorithms.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1530-437X
1558-1748
DOI:10.1109/JSEN.2024.3444053