AN ADAPTIVE RESONANCE THEORY-BASED ARTIFICIAL NEURAL-NETWORK (ART-2A) FOR RAPID IDENTIFICATION OF AIRBORNE PARTICLE SHAPES FROM THEIR SCANNING-ELECTRON-MICROSCOPY IMAGES

An adaptive resonance theory based artificial neural network (ART-2a) is applied to the rapid identification of distinct shapes of airborne particles. Feature vectors describing the particle shapes have been extracted from binary-coded 8-bit grey level scanning electron microscopy images of particles. This shape information has been combined with feature vectors describing the chemical composition of airborne particles. After feature selection and preprocessing, the distinct particle shapes were classified by an ART-2a type network. Additionally the interclass angle has been introduced for network diagnostics. The network is robust against outliers and new classes of sample vectors that were not included into the training set.