Automatic Identification of Lyocell and Cotton Fibers Using Cluster Analysis

Abstract

Lyocell is a cellulosic fiber manufactured through a more environmentally friendly process. As lyocell and cotton have many complementary properties, such as bending elasticity, deliquescent effect and gliding property, they are often blended for use in dresses and formal wear for better comfort and drape. The quantitative analysis of lyocell and cotton blends is important for manufacturers, dealers, retailers, service providers or suppliers of blended textiles, and for independent testing labs.In our previous research, an automatic system for identifying ramie and cotton fibers was designed [1,2]. The system can automatically capture a series of microscopic fiber images in the longitudinal view to form a panoramic image of long-fiber segments, and then locate each valid frame to capture well-focused fiber images. Six characteristic parameters based on the Y-projection and X-projections of fiber stripes [1-4] were established for the fiber classification. These parameters, denoted as P i (i = 1~6), are listed in Table 1. The correlation analysis of the six parameters for ramie and cotton fibers revealed that they are linearly dependent [2]. 1 In this study, we will expand the function of this system for lyocell and cotton fiber identifications. Since the image capturing and feature extraction methods remain the same, we will focus on more effective pattern recognition Abstract This study applies cluster analysis to identifying two frequently blended fibers, lyocell and cotton, based on the six characteristic parameters extracted from the automatic fiber identification system presented in the previous publications. Two independent parameters are first derived from the six characteristic parameters by using factor analysis. Second, a probability density distribution map of the two indirect parameters is established through sample observations. Finally, the clusters of lyocell and cotton fibers in the probability density distribution map are segmented according to contour lines and distance. The experiment showed that the accuracy of lyocell and cotton fiber identification with the cluster analysis is above 95%. Table 1 Definitions of six fiber parameters.P 1 The average width of a fiber. P 2 The number of wave crests in the fluctuant fiber-width curve with values over 25% of the average fiber width in the range of one millimeter. P 3 The projective width of the X projection.P 4 The average of the subpoints on the Y projection.P 5 The number of the wave crests in the X projection.P 6 The depth of the deepest trough in the X projection.

An improved algorithm for three-dimensional reconstruction of nonwoven microstructures by using multi-focus images

“…Actually, the identification of fibers with similar morphology has always been one of the essential topics in the field of textiles. Wang et al 19 for example, accomplished the segmentation algorithm of lyocell and cotton fibers by analyzing the feature factors. Zhang et al 20 implemented the identification of overlapped wool/cashmere fibers using a convolution neural network (CNN).…”

“…Wang et al. 19 for example, accomplished the segmentation algorithm of lyocell and cotton fibers by analyzing the feature factors. Zhang et al.…”