TAOS (two-angle optical scattering) is a well-known experimental technique which records the intensity patterns of laser light scattered by single aerosol particles over an extended range of the scattering angles {θ, φ} [1]. In the absence of a method which solves the inverse obstacle problem from intensity data, patterns have to be classified by artificial intelligence tech- niques. The classifier described herewith extracts features from TAOS patterns by the spectrum enhancement algorithm [2], which is controlled by a few parameters (the n-tuple ψ), not described herewith. Training corresponds to finding a ψ which maximizes a suitable figure of merit (F ). The newly developed training-validation scheme is illustrated by the following example. 1) Out of 100 TAOS patterns produced by clusters of polystyrene spheres of controlled size (material a6, class 1) and 100 patterns from single spores of Bacil lus globigii (material bg, class 2), form e.g., 10 training (T ) sets of 50 patterns from each class, which differ by at least 5 + 5 patterns from one another. Two sample patterns are shown by Figures 1 and 2 below
Crosta, G., Pan, Y., Fernandes, G. (2009). Training and Validation of a Wide-angle Optical Scattering (TAOS) Pattern Classifier. In PIERS 2009 Moscow Progress In Electromagnetics Research Symposium Draft Abstracts (pp.558-559). Cambridge, MA : The Electromagnetics Academy.
Training and Validation of a Wide-angle Optical Scattering (TAOS) Pattern Classifier
CROSTA, GIOVANNI FRANCO FILIPPO;
2009
Abstract
TAOS (two-angle optical scattering) is a well-known experimental technique which records the intensity patterns of laser light scattered by single aerosol particles over an extended range of the scattering angles {θ, φ} [1]. In the absence of a method which solves the inverse obstacle problem from intensity data, patterns have to be classified by artificial intelligence tech- niques. The classifier described herewith extracts features from TAOS patterns by the spectrum enhancement algorithm [2], which is controlled by a few parameters (the n-tuple ψ), not described herewith. Training corresponds to finding a ψ which maximizes a suitable figure of merit (F ). The newly developed training-validation scheme is illustrated by the following example. 1) Out of 100 TAOS patterns produced by clusters of polystyrene spheres of controlled size (material a6, class 1) and 100 patterns from single spores of Bacil lus globigii (material bg, class 2), form e.g., 10 training (T ) sets of 50 patterns from each class, which differ by at least 5 + 5 patterns from one another. Two sample patterns are shown by Figures 1 and 2 belowFile | Dimensione | Formato | |
---|---|---|---|
Training_and_Validation_Wide-angle_Optical_Scattering_Abstract.pdf
accesso aperto
Tipologia di allegato:
Other attachments
Dimensione
212.23 kB
Formato
Adobe PDF
|
212.23 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.