Evaluation of yarn characteristics using computer vision and image processing

Irregularity, hairiness and twist are among the most important characteristics that define yarn quality. This thesis describes computer vision and image processing techniques developed to evaluate these characteristics. The optical and electronic aspects such as the illumination, lens parameters and aberrations play crucial role on the quality of yam images and on the overall performance of image processing. The depth of field limitation being the most important restraint in yam imaging as well as image distortion in line scan cameras arising from digitisation and yam movement are modelled mathematically and verified through experiments both for front-lit and back-lit illuminations. Various light sources and arrangements are tested and relative advantages and disadvantages are discussed based on the image quality. Known problems in defining the hair-core boundaries and determining the total hairiness from yam images are addressed and image enhancement and processing algorithms developed to overcome these problems are explained. A method to simulate various yam scanning resolution conditions is described. Using this method, the minimum scanning resolution limits to measure the hairiness and irregularity are investigated. [Continues.]