Design and implementation of a biological warfare agents simulant spectral database for the automatic discrimination in unconventional scenarios

The biological agents (BAs) can exist in the form of aerosolized bacteria, fungi or viruses, and they also may be spread by vectors and other routes like water and air supply (Cenciarelli et al., 2013; Kumar et al., 2013). Among the ways which allow the propagation of BAs, certainly air has a more rapid rate of diffusion; for this reason it is important to develop systems for remote sensing of airborne biological agents. Jonsson et al. (2005) successfully present a detection system for BAs based on spectral detection of ultraviolet (UV) laser induced fluorescence (LIF); spectral characterization was also illustrated for some simulants. The most common fluorophores found in bioaerosols are phenylalanine, tyrosine, tryptophan, nicotinamide adenine dinucleotide compounds (NADH), flavins and chlorophylls. Among these main fluorophores in bioaerosols, phenylalanine has little absorption above 270 nm, while chlorophylls adsorb around 430 nm. Excitation around 260-290 nm will mainly generate fluorescence emission from tyrosine, tryptophan, NADH, and flavin, while excitation around 350-410 nm will mainly produce fluorescence emission from NADH and flavin. Although in the literature several studies regarding the ability to discriminate BAs through methods that exploit fluorescence excitation through ultraviolet radiation were presented, it appears to be essential to build and implement a database being able to return a response about the kind of BA detected according to a revealed fluorescence spectrum. Furthermore, in order to refine the experimental procedures, it appears essential to standardize the samples preparation methods. This represents the starting point for the creation of a pitched instrument. This PhD research activity aims to design a biological warfare agents (BWA) spectral database using biological agents simulants (BWA-S) for the automatic discrimination in unconventional scenarios, i.e. by standardizing the experimental procedures (e.g. best samples preparation) and setting the better criteria for the algorithm analysis. The PhD in Industrial Engineering - Final Dissertation - Orlando Cenciarelli DESIGN AND IMPLEMENTATION OF A BIOLOGICAL WARFARE AGENTS SIMULANT SPECTRAL DATABASE FOR THE AUTOMATIC DISCRIMINATION IN UNCONVENTIONAL SCENARIOS 3 automatic classification system is based on a novel adaptative algorithm, named as Universal Multi-Event Locator (UMEL). UMEL is a universal technique based on Support Vector Regression (SVR), a version of Support Vector Machines (SVM) for function estimation (Carestia et al., 2014). SVR fits the training data discarding factors such as sampling rate or noise distribution. This technique computes a fitting function and, in addition, retrieves a list of points from the training set. These points become Support Vectors (SVs). For the first time, UMEL is used in this work to discriminate between fluorescence spectral signature of BWA-S.