Fiber Recognition Of Cotton/Rayon Blended Yarn Based On Image Analysis

The method of chemical dissolve-to-weigh has often been used to detect yarn blending ratio for a long time. The method can not effectively work, when blended yarns are composed of two or more kinds of fibers having similar chemical characteristics (for example: cotton/rayon, wool/cashmere). Image processing technique opened the way for the identification of blending ratio of these blended yarns. So, the research on it has practicability and significance. Taking cotton/rayon blended yarn for experimental subject and summarizing the study of predecessors, this paper deals with the details in the five aspects: image capturing, image preprocessing, individual detection, feature extraction and cluster analysis of feature data–which compose the whole process of this research..On image capturing, epoxy resin embedding technique is applied to get high quality, little deformation of fiber cross-section, dispersed highly slice images. It is the better foundation for the following image analysis.On image preprocessing, these slice images have been processed by using opening operation, closing operation, image reconstruction, and image enhance of mathematical morphology. It has concretely investigated the problem how to select parameters of morphology operation and has raised the preprocessing solution for slice images of this paper.On individual detection, Facula Diffusion Model is used. The eye sight principle of the model and control parameters of facula diffusion are expounded. By improving its single-pot diffusion mode, it has solved the problem that highly crimped cross-section fibers can not been completely detected in the diffusion mode.On feature extraction and cluster analysis, seven shape indices are proposed which include Deviation, Fluctuation, Abnormity and so on. According to the span distribution character, the Fluctuate which can stand for the distinguish characteristics between cotton and rayon fiber are designed. The seven shape indices have separately been used to sample clustering analysis by Fuzzy C-means cluster algorithm and Hierarchical clustering analysis. Experiment results show that using Fluctuate can both distinguish cotton and rayon fiber. Compared with Hierarchical clustering analysis method, Fuzzy C-means cluster algorithm represents a significant advance and can be applied to fiber recognition more effectively.