AAAR 35th Annual Conference October 17 - October 21, 2016 Oregon Convention Center Portland, Oregon, USA
Abstract View
Classification of Airborne Particulates Using Multispectral Light Scattering Imaging
STEPHEN HOLLER, Stephen Fuerstenau, Charles Skelsey, Fordham University
Abstract Number: 611 Working Group: Single Aerosol Particle Studies - Techniques and Instrumentation
Abstract Light scattering patterns from non-spherical particles and aggregates exhibit complex structure that is only revealed by observing the scattering pattern over a large solid angle. Because of morphological differences in these aerosols, whether from structure or composition, the rich structure in the two-dimensional angular optical scattering (TAOS) patterns vary from particle to particle, even when they belong to the same class. We have investigated such TAOS patterns using a single color CCD camera to simultaneously capture spectrally distinct patterns from individual aerosols. Since the optical size of the scattering particle is inversely proportional to the illuminating wavelength, the spectrally resolved scattering information provides characteristic information about the airborne particles simultaneously in two different scaling regimes. The simultaneous acquisition of data from airborne particulate matter at two different wavelengths allows for additional degrees of freedom in the analysis and characterization of the aerosols. We examine two-dimensional light scattering patterns obtained at multiple wavelengths using a single CCD camera with minimal cross talk between channels. The integration of the approach with a single CCD camera assures that data is acquired within the same solid angle and orientation. We have attempted to deconstruct the recorded TAOS patterns in order to identify characteristic descriptors that can describe the morphology and be implemented in multivariate algorithms for classification. The results of these classifications indicate that the descriptors we developed are meaningful for testing classes against a known database. The classifications were tested against several metrics, and we have achieved classification on par with previous analyses, but with morphological descriptors that reflect physical characteristics of the aerosols.