Forensic Textile Fibre Matching by Multiresolution Planning and Colour Space Pattern Recognition

Colour is a key element in the discrimination and matching of textile fibres for forensic purposes. Two aspects of research aimed at developing an autcmated forensic textile fibre matching system by colour computer vision are outlined. The segnentation procedure is described. It is based on a mult{resolution image format and incorporates progressive plan refinement. Filters for building the milt {resolution image are compared and reconwendations made as to when they might be useful. The use of colour, as it relates to this application, is formulated in terms of classical pattern recognition. The colour space clusters associated with textile fibre classes are described and a suitable clustering scheme is presented.