Classification of wood pulp fibre cross-sectional shapes

Abstract

This work presents a comparison of two statistical approaches for automatic classification of fibre shapes, i.e. Canonical Discriminant Analysis (CDA) and Mahalanobis Discriminant Analysis (MLDA). The discriminant analyses were applied to identify and classify several fibre cross-sectional shapes, including e.g. intact, collapsed, touching and fibrillated fibres. The discriminant analyses perform differently, giving clear indications of their suitability for classifying a given group of fibre elements. Compared to CDA, MLDA was more reliable and relatively stable. © 2010 Springer-Verlag.