Airborne material particles in the 5 micromolar size range have been collected, resuspended and analyzed by the TAOS (two-dimensional angular optical scattering) technique. The corresponding patterns of light intensity scattered by single particles have been automatically classified by an algorithm based on "spectrum enhancement", multivariate statistics and supervised optimization. The enhanced spectrum has resulted from some non-linear operations on fractional spatial derivatives of the pattern. It has yielded morphological descriptors of the pattern. A multiobjective optimization algorithm has included principal components analysis and has maximized pairwise discrimination between classes. The classifier has been trained by TAOS patterns from 10 micromolar polystyrene spheres (P) and background aerosol particles (B). Then it has been applied to recognize patterns from airborne debris (A) sampled on a car racing track. Training with at least 10 patterns per class has discriminated P and B from A at confidence levels >or=90%. Training by samples of smaller sizes (e.g., 5P and 12B patterns) has obviously yielded lower confidence levels (65% in B-A discrimination).

Crosta, G. (2005). Classification of single particle optical scattering patterns by the spectrum enhancement algorithm. In A. III Sedlacek, S. Christesen, R. Combs, T. Vo-Dinh (a cura di), Chemical and Biological Sensors for Industrial and Environmental Security (pp. 599402-1-599402-12). Bellingham, WA : SPIE-INT SOCIETY OPTICAL ENGINEERING [10.1117/12.629064].

Classification of single particle optical scattering patterns by the spectrum enhancement algorithm

CROSTA, GIOVANNI FRANCO FILIPPO
2005

Abstract

Airborne material particles in the 5 micromolar size range have been collected, resuspended and analyzed by the TAOS (two-dimensional angular optical scattering) technique. The corresponding patterns of light intensity scattered by single particles have been automatically classified by an algorithm based on "spectrum enhancement", multivariate statistics and supervised optimization. The enhanced spectrum has resulted from some non-linear operations on fractional spatial derivatives of the pattern. It has yielded morphological descriptors of the pattern. A multiobjective optimization algorithm has included principal components analysis and has maximized pairwise discrimination between classes. The classifier has been trained by TAOS patterns from 10 micromolar polystyrene spheres (P) and background aerosol particles (B). Then it has been applied to recognize patterns from airborne debris (A) sampled on a car racing track. Training with at least 10 patterns per class has discriminated P and B from A at confidence levels >or=90%. Training by samples of smaller sizes (e.g., 5P and 12B patterns) has obviously yielded lower confidence levels (65% in B-A discrimination).
Capitolo o saggio
aerosols; feature extraction; light scattering; optimisation; particle size measurement; pattern classification; polymers; principal component analysis
English
Chemical and Biological Sensors for Industrial and Environmental Security
III Sedlacek, A.J.; Christesen, S.D.; Combs, R.J.; Vo-Dinh, T.
2005
0-8194-6018-4
SPIE-INT SOCIETY OPTICAL ENGINEERING
599402-1
599402-12
Crosta, G. (2005). Classification of single particle optical scattering patterns by the spectrum enhancement algorithm. In A. III Sedlacek, S. Christesen, R. Combs, T. Vo-Dinh (a cura di), Chemical and Biological Sensors for Industrial and Environmental Security (pp. 599402-1-599402-12). Bellingham, WA : SPIE-INT SOCIETY OPTICAL ENGINEERING [10.1117/12.629064].
none
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/4004
Citazioni
  • Scopus 6
  • ???jsp.display-item.citation.isi??? ND
Social impact